Ghost/ghost/email-analytics-service/lib/EmailAnalyticsService.js

428 lines
16 KiB
JavaScript
Raw Normal View History

const EventProcessingResult = require('./EventProcessingResult');
const logging = require('@tryghost/logging');
Implemented email analytics retrying (#16273) fixes https://github.com/TryGhost/Team/issues/2562 New event fetching loops: - Reworked the analytics fetching algorithm. Instead of starting again where we stopped during the last fetching minus 30 minutes, we now just continue where we stopped. But with ms precision (because no longer database dependent after first fetch), and we stop at NOW - 1 minute to reduce chance of missing events. - Apart from that, a missing fetching loop is introduced. This fetches events that are older than 30 minutes, and just processes all events a second time to make sure we didn't skip any because of storage delays in the Mailgun API. - A new scheduled fetching loop, that allows us to schedule between a given start/end date (currently only persisted in memory, so stops after a reboot) UI and endpoint changes: - New UI to show the state of the analytics 'loops' - New endpoint to request the analytics loop status - New endpoint to schedule analytics - New endpoint to cancel scheduled analytics - Some number formatting improvements, and introduction of 'opened' count in debug screen - Live reload of data in the debug screen Other changes: - This also improves the support for maxEvents. We can now stop a fetching loop after x events without worrying about lost events. This is used to reduce the fetched events in the missing and scheduled event loop (e.g. when the main one is fetching lots of events, we skip the other loops). - Prevents fetching the same events over and over again if no new events come in (because we always started at the same begin timestamp). The code increases the begin timestamp with 1 second if it is safe to do so, to prevent the API from returning the same events over and over again. - Some optimisations in handing the processing results (less merges to reduce CPU usage in cases we have lots of events). Testing: - You can test with lots of events using the new mailgun mocking server (Toolbox repo `scripts/mailgun-mock-server`). This can also simulate events that are only returned after x minutes because of storage delays.
2023-02-20 18:44:13 +03:00
const errors = require('@tryghost/errors');
Added new email event processor (#15879) fixes https://github.com/TryGhost/Team/issues/2310 This moves the processing of the events from the event-processor to a new email-event-processor in the email-service package. - The `EmailEventProcessor` only translates events from providerId/emailId to their known emailId, memberId and recipientId, and dispatches the corresponding events. - Since `EmailEventProcessor` runs in a separate worker thread, we can't listen for the dispatched events on the main thread. To accomplish this communication, the events dispatched from the `EmailEventProcessor` class are 'posted' via the postMessage method and redispatched on the main thread. - A new `EmailEventStorage` class reacts to the email events and stores it in the database. This code mostly corresponds to the (now deleted) subclass of the old `EmailEventProcessor` - Updating a members last_seen_at timestamp has moved to the lastSeenAtUpdater. - Email events no longer store `ObjectID` because these are not encodable across threads via postMessage - Includes new E2E tests that test the storage of all supported Mailgun events. Note that in these tests we run the processing on the main thread instead of on a separate thread (couldn't do this because stubbing is not possible across threads) There are some missing pieces that will get added in later PRs (this PR focuses on porting the existing functionality): - Handling temporary failures/bounces - Capturing the error messages of bounce events
2022-11-29 13:15:19 +03:00
/**
* @typedef {import('@tryghost/email-service').EmailEventProcessor} EmailEventProcessor
*/
Implemented email analytics retrying (#16273) fixes https://github.com/TryGhost/Team/issues/2562 New event fetching loops: - Reworked the analytics fetching algorithm. Instead of starting again where we stopped during the last fetching minus 30 minutes, we now just continue where we stopped. But with ms precision (because no longer database dependent after first fetch), and we stop at NOW - 1 minute to reduce chance of missing events. - Apart from that, a missing fetching loop is introduced. This fetches events that are older than 30 minutes, and just processes all events a second time to make sure we didn't skip any because of storage delays in the Mailgun API. - A new scheduled fetching loop, that allows us to schedule between a given start/end date (currently only persisted in memory, so stops after a reboot) UI and endpoint changes: - New UI to show the state of the analytics 'loops' - New endpoint to request the analytics loop status - New endpoint to schedule analytics - New endpoint to cancel scheduled analytics - Some number formatting improvements, and introduction of 'opened' count in debug screen - Live reload of data in the debug screen Other changes: - This also improves the support for maxEvents. We can now stop a fetching loop after x events without worrying about lost events. This is used to reduce the fetched events in the missing and scheduled event loop (e.g. when the main one is fetching lots of events, we skip the other loops). - Prevents fetching the same events over and over again if no new events come in (because we always started at the same begin timestamp). The code increases the begin timestamp with 1 second if it is safe to do so, to prevent the API from returning the same events over and over again. - Some optimisations in handing the processing results (less merges to reduce CPU usage in cases we have lots of events). Testing: - You can test with lots of events using the new mailgun mocking server (Toolbox repo `scripts/mailgun-mock-server`). This can also simulate events that are only returned after x minutes because of storage delays.
2023-02-20 18:44:13 +03:00
/**
* @typedef {object} FetchData
* @property {boolean} running
* @property {Date} [lastStarted] Date the last fetch started on
* @property {Date} [lastBegin] The begin time used during the last fetch
* @property {Date} [lastEventTimestamp]
* @property {boolean} [canceled] Set to quit the job early
*/
/**
* @typedef {FetchData & {schedule: {begin: Date, end: Date}}} FetchDataScheduled
*/
const TRUST_THRESHOLD_MS = 30 * 60 * 1000; // 30 minutes
const FETCH_LATEST_END_MARGIN_MS = 1 * 60 * 1000; // Do not fetch events newer than 1 minute (yet). Reduces the chance of having missed events in fetchLatest.
module.exports = class EmailAnalyticsService {
Added new email event processor (#15879) fixes https://github.com/TryGhost/Team/issues/2310 This moves the processing of the events from the event-processor to a new email-event-processor in the email-service package. - The `EmailEventProcessor` only translates events from providerId/emailId to their known emailId, memberId and recipientId, and dispatches the corresponding events. - Since `EmailEventProcessor` runs in a separate worker thread, we can't listen for the dispatched events on the main thread. To accomplish this communication, the events dispatched from the `EmailEventProcessor` class are 'posted' via the postMessage method and redispatched on the main thread. - A new `EmailEventStorage` class reacts to the email events and stores it in the database. This code mostly corresponds to the (now deleted) subclass of the old `EmailEventProcessor` - Updating a members last_seen_at timestamp has moved to the lastSeenAtUpdater. - Email events no longer store `ObjectID` because these are not encodable across threads via postMessage - Includes new E2E tests that test the storage of all supported Mailgun events. Note that in these tests we run the processing on the main thread instead of on a separate thread (couldn't do this because stubbing is not possible across threads) There are some missing pieces that will get added in later PRs (this PR focuses on porting the existing functionality): - Handling temporary failures/bounces - Capturing the error messages of bounce events
2022-11-29 13:15:19 +03:00
config;
settings;
queries;
eventProcessor;
providers;
Implemented email analytics retrying (#16273) fixes https://github.com/TryGhost/Team/issues/2562 New event fetching loops: - Reworked the analytics fetching algorithm. Instead of starting again where we stopped during the last fetching minus 30 minutes, we now just continue where we stopped. But with ms precision (because no longer database dependent after first fetch), and we stop at NOW - 1 minute to reduce chance of missing events. - Apart from that, a missing fetching loop is introduced. This fetches events that are older than 30 minutes, and just processes all events a second time to make sure we didn't skip any because of storage delays in the Mailgun API. - A new scheduled fetching loop, that allows us to schedule between a given start/end date (currently only persisted in memory, so stops after a reboot) UI and endpoint changes: - New UI to show the state of the analytics 'loops' - New endpoint to request the analytics loop status - New endpoint to schedule analytics - New endpoint to cancel scheduled analytics - Some number formatting improvements, and introduction of 'opened' count in debug screen - Live reload of data in the debug screen Other changes: - This also improves the support for maxEvents. We can now stop a fetching loop after x events without worrying about lost events. This is used to reduce the fetched events in the missing and scheduled event loop (e.g. when the main one is fetching lots of events, we skip the other loops). - Prevents fetching the same events over and over again if no new events come in (because we always started at the same begin timestamp). The code increases the begin timestamp with 1 second if it is safe to do so, to prevent the API from returning the same events over and over again. - Some optimisations in handing the processing results (less merges to reduce CPU usage in cases we have lots of events). Testing: - You can test with lots of events using the new mailgun mocking server (Toolbox repo `scripts/mailgun-mock-server`). This can also simulate events that are only returned after x minutes because of storage delays.
2023-02-20 18:44:13 +03:00
/**
* @type {FetchData}
*/
#fetchLatestData = null;
/**
* @type {FetchData}
*/
#fetchMissingData = null;
/**
* @type {FetchDataScheduled}
*/
#fetchScheduledData = null;
Added new email event processor (#15879) fixes https://github.com/TryGhost/Team/issues/2310 This moves the processing of the events from the event-processor to a new email-event-processor in the email-service package. - The `EmailEventProcessor` only translates events from providerId/emailId to their known emailId, memberId and recipientId, and dispatches the corresponding events. - Since `EmailEventProcessor` runs in a separate worker thread, we can't listen for the dispatched events on the main thread. To accomplish this communication, the events dispatched from the `EmailEventProcessor` class are 'posted' via the postMessage method and redispatched on the main thread. - A new `EmailEventStorage` class reacts to the email events and stores it in the database. This code mostly corresponds to the (now deleted) subclass of the old `EmailEventProcessor` - Updating a members last_seen_at timestamp has moved to the lastSeenAtUpdater. - Email events no longer store `ObjectID` because these are not encodable across threads via postMessage - Includes new E2E tests that test the storage of all supported Mailgun events. Note that in these tests we run the processing on the main thread instead of on a separate thread (couldn't do this because stubbing is not possible across threads) There are some missing pieces that will get added in later PRs (this PR focuses on porting the existing functionality): - Handling temporary failures/bounces - Capturing the error messages of bounce events
2022-11-29 13:15:19 +03:00
/**
* @param {object} dependencies
Added new email event processor (#15879) fixes https://github.com/TryGhost/Team/issues/2310 This moves the processing of the events from the event-processor to a new email-event-processor in the email-service package. - The `EmailEventProcessor` only translates events from providerId/emailId to their known emailId, memberId and recipientId, and dispatches the corresponding events. - Since `EmailEventProcessor` runs in a separate worker thread, we can't listen for the dispatched events on the main thread. To accomplish this communication, the events dispatched from the `EmailEventProcessor` class are 'posted' via the postMessage method and redispatched on the main thread. - A new `EmailEventStorage` class reacts to the email events and stores it in the database. This code mostly corresponds to the (now deleted) subclass of the old `EmailEventProcessor` - Updating a members last_seen_at timestamp has moved to the lastSeenAtUpdater. - Email events no longer store `ObjectID` because these are not encodable across threads via postMessage - Includes new E2E tests that test the storage of all supported Mailgun events. Note that in these tests we run the processing on the main thread instead of on a separate thread (couldn't do this because stubbing is not possible across threads) There are some missing pieces that will get added in later PRs (this PR focuses on porting the existing functionality): - Handling temporary failures/bounces - Capturing the error messages of bounce events
2022-11-29 13:15:19 +03:00
* @param {EmailEventProcessor} dependencies.eventProcessor
*/
constructor({config, settings, queries, eventProcessor, providers}) {
this.config = config;
this.settings = settings;
this.queries = queries;
this.eventProcessor = eventProcessor;
this.providers = providers;
}
Implemented email analytics retrying (#16273) fixes https://github.com/TryGhost/Team/issues/2562 New event fetching loops: - Reworked the analytics fetching algorithm. Instead of starting again where we stopped during the last fetching minus 30 minutes, we now just continue where we stopped. But with ms precision (because no longer database dependent after first fetch), and we stop at NOW - 1 minute to reduce chance of missing events. - Apart from that, a missing fetching loop is introduced. This fetches events that are older than 30 minutes, and just processes all events a second time to make sure we didn't skip any because of storage delays in the Mailgun API. - A new scheduled fetching loop, that allows us to schedule between a given start/end date (currently only persisted in memory, so stops after a reboot) UI and endpoint changes: - New UI to show the state of the analytics 'loops' - New endpoint to request the analytics loop status - New endpoint to schedule analytics - New endpoint to cancel scheduled analytics - Some number formatting improvements, and introduction of 'opened' count in debug screen - Live reload of data in the debug screen Other changes: - This also improves the support for maxEvents. We can now stop a fetching loop after x events without worrying about lost events. This is used to reduce the fetched events in the missing and scheduled event loop (e.g. when the main one is fetching lots of events, we skip the other loops). - Prevents fetching the same events over and over again if no new events come in (because we always started at the same begin timestamp). The code increases the begin timestamp with 1 second if it is safe to do so, to prevent the API from returning the same events over and over again. - Some optimisations in handing the processing results (less merges to reduce CPU usage in cases we have lots of events). Testing: - You can test with lots of events using the new mailgun mocking server (Toolbox repo `scripts/mailgun-mock-server`). This can also simulate events that are only returned after x minutes because of storage delays.
2023-02-20 18:44:13 +03:00
getStatus() {
return {
latest: this.#fetchLatestData,
missing: this.#fetchMissingData,
scheduled: this.#fetchScheduledData
};
}
/**
* Returns the timestamp of the last event we processed. Defaults to now minus 30 minutes if we have no data yet.
*/
async getLastEventTimestamp() {
return this.#fetchLatestData?.lastEventTimestamp ?? (await this.queries.getLastSeenEventTimestamp()) ?? new Date(Date.now() - TRUST_THRESHOLD_MS);
}
Implemented email analytics retrying (#16273) fixes https://github.com/TryGhost/Team/issues/2562 New event fetching loops: - Reworked the analytics fetching algorithm. Instead of starting again where we stopped during the last fetching minus 30 minutes, we now just continue where we stopped. But with ms precision (because no longer database dependent after first fetch), and we stop at NOW - 1 minute to reduce chance of missing events. - Apart from that, a missing fetching loop is introduced. This fetches events that are older than 30 minutes, and just processes all events a second time to make sure we didn't skip any because of storage delays in the Mailgun API. - A new scheduled fetching loop, that allows us to schedule between a given start/end date (currently only persisted in memory, so stops after a reboot) UI and endpoint changes: - New UI to show the state of the analytics 'loops' - New endpoint to request the analytics loop status - New endpoint to schedule analytics - New endpoint to cancel scheduled analytics - Some number formatting improvements, and introduction of 'opened' count in debug screen - Live reload of data in the debug screen Other changes: - This also improves the support for maxEvents. We can now stop a fetching loop after x events without worrying about lost events. This is used to reduce the fetched events in the missing and scheduled event loop (e.g. when the main one is fetching lots of events, we skip the other loops). - Prevents fetching the same events over and over again if no new events come in (because we always started at the same begin timestamp). The code increases the begin timestamp with 1 second if it is safe to do so, to prevent the API from returning the same events over and over again. - Some optimisations in handing the processing results (less merges to reduce CPU usage in cases we have lots of events). Testing: - You can test with lots of events using the new mailgun mocking server (Toolbox repo `scripts/mailgun-mock-server`). This can also simulate events that are only returned after x minutes because of storage delays.
2023-02-20 18:44:13 +03:00
async fetchLatest({maxEvents = Infinity} = {}) {
// Start where we left of, or the last stored event in the database, or start 30 minutes ago if we have nothing available
const begin = await this.getLastEventTimestamp();
const end = new Date(Date.now() - FETCH_LATEST_END_MARGIN_MS); // ALways stop at x minutes ago to give Mailgun a bit more time to stabilize storage
if (end <= begin) {
Implemented email analytics retrying (#16273) fixes https://github.com/TryGhost/Team/issues/2562 New event fetching loops: - Reworked the analytics fetching algorithm. Instead of starting again where we stopped during the last fetching minus 30 minutes, we now just continue where we stopped. But with ms precision (because no longer database dependent after first fetch), and we stop at NOW - 1 minute to reduce chance of missing events. - Apart from that, a missing fetching loop is introduced. This fetches events that are older than 30 minutes, and just processes all events a second time to make sure we didn't skip any because of storage delays in the Mailgun API. - A new scheduled fetching loop, that allows us to schedule between a given start/end date (currently only persisted in memory, so stops after a reboot) UI and endpoint changes: - New UI to show the state of the analytics 'loops' - New endpoint to request the analytics loop status - New endpoint to schedule analytics - New endpoint to cancel scheduled analytics - Some number formatting improvements, and introduction of 'opened' count in debug screen - Live reload of data in the debug screen Other changes: - This also improves the support for maxEvents. We can now stop a fetching loop after x events without worrying about lost events. This is used to reduce the fetched events in the missing and scheduled event loop (e.g. when the main one is fetching lots of events, we skip the other loops). - Prevents fetching the same events over and over again if no new events come in (because we always started at the same begin timestamp). The code increases the begin timestamp with 1 second if it is safe to do so, to prevent the API from returning the same events over and over again. - Some optimisations in handing the processing results (less merges to reduce CPU usage in cases we have lots of events). Testing: - You can test with lots of events using the new mailgun mocking server (Toolbox repo `scripts/mailgun-mock-server`). This can also simulate events that are only returned after x minutes because of storage delays.
2023-02-20 18:44:13 +03:00
// Skip for now
logging.info('[EmailAnalytics] Skipping fetchLatest because end (' + end + ') is before begin (' + begin + ')');
return 0;
}
Implemented email analytics retrying (#16273) fixes https://github.com/TryGhost/Team/issues/2562 New event fetching loops: - Reworked the analytics fetching algorithm. Instead of starting again where we stopped during the last fetching minus 30 minutes, we now just continue where we stopped. But with ms precision (because no longer database dependent after first fetch), and we stop at NOW - 1 minute to reduce chance of missing events. - Apart from that, a missing fetching loop is introduced. This fetches events that are older than 30 minutes, and just processes all events a second time to make sure we didn't skip any because of storage delays in the Mailgun API. - A new scheduled fetching loop, that allows us to schedule between a given start/end date (currently only persisted in memory, so stops after a reboot) UI and endpoint changes: - New UI to show the state of the analytics 'loops' - New endpoint to request the analytics loop status - New endpoint to schedule analytics - New endpoint to cancel scheduled analytics - Some number formatting improvements, and introduction of 'opened' count in debug screen - Live reload of data in the debug screen Other changes: - This also improves the support for maxEvents. We can now stop a fetching loop after x events without worrying about lost events. This is used to reduce the fetched events in the missing and scheduled event loop (e.g. when the main one is fetching lots of events, we skip the other loops). - Prevents fetching the same events over and over again if no new events come in (because we always started at the same begin timestamp). The code increases the begin timestamp with 1 second if it is safe to do so, to prevent the API from returning the same events over and over again. - Some optimisations in handing the processing results (less merges to reduce CPU usage in cases we have lots of events). Testing: - You can test with lots of events using the new mailgun mocking server (Toolbox repo `scripts/mailgun-mock-server`). This can also simulate events that are only returned after x minutes because of storage delays.
2023-02-20 18:44:13 +03:00
// Create the fetch data object if it doesn't exist yet
if (!this.#fetchLatestData) {
this.#fetchLatestData = {
running: false
};
}
Implemented email analytics retrying (#16273) fixes https://github.com/TryGhost/Team/issues/2562 New event fetching loops: - Reworked the analytics fetching algorithm. Instead of starting again where we stopped during the last fetching minus 30 minutes, we now just continue where we stopped. But with ms precision (because no longer database dependent after first fetch), and we stop at NOW - 1 minute to reduce chance of missing events. - Apart from that, a missing fetching loop is introduced. This fetches events that are older than 30 minutes, and just processes all events a second time to make sure we didn't skip any because of storage delays in the Mailgun API. - A new scheduled fetching loop, that allows us to schedule between a given start/end date (currently only persisted in memory, so stops after a reboot) UI and endpoint changes: - New UI to show the state of the analytics 'loops' - New endpoint to request the analytics loop status - New endpoint to schedule analytics - New endpoint to cancel scheduled analytics - Some number formatting improvements, and introduction of 'opened' count in debug screen - Live reload of data in the debug screen Other changes: - This also improves the support for maxEvents. We can now stop a fetching loop after x events without worrying about lost events. This is used to reduce the fetched events in the missing and scheduled event loop (e.g. when the main one is fetching lots of events, we skip the other loops). - Prevents fetching the same events over and over again if no new events come in (because we always started at the same begin timestamp). The code increases the begin timestamp with 1 second if it is safe to do so, to prevent the API from returning the same events over and over again. - Some optimisations in handing the processing results (less merges to reduce CPU usage in cases we have lots of events). Testing: - You can test with lots of events using the new mailgun mocking server (Toolbox repo `scripts/mailgun-mock-server`). This can also simulate events that are only returned after x minutes because of storage delays.
2023-02-20 18:44:13 +03:00
return await this.#fetchEvents(this.#fetchLatestData, {begin, end, maxEvents});
}
Implemented email analytics retrying (#16273) fixes https://github.com/TryGhost/Team/issues/2562 New event fetching loops: - Reworked the analytics fetching algorithm. Instead of starting again where we stopped during the last fetching minus 30 minutes, we now just continue where we stopped. But with ms precision (because no longer database dependent after first fetch), and we stop at NOW - 1 minute to reduce chance of missing events. - Apart from that, a missing fetching loop is introduced. This fetches events that are older than 30 minutes, and just processes all events a second time to make sure we didn't skip any because of storage delays in the Mailgun API. - A new scheduled fetching loop, that allows us to schedule between a given start/end date (currently only persisted in memory, so stops after a reboot) UI and endpoint changes: - New UI to show the state of the analytics 'loops' - New endpoint to request the analytics loop status - New endpoint to schedule analytics - New endpoint to cancel scheduled analytics - Some number formatting improvements, and introduction of 'opened' count in debug screen - Live reload of data in the debug screen Other changes: - This also improves the support for maxEvents. We can now stop a fetching loop after x events without worrying about lost events. This is used to reduce the fetched events in the missing and scheduled event loop (e.g. when the main one is fetching lots of events, we skip the other loops). - Prevents fetching the same events over and over again if no new events come in (because we always started at the same begin timestamp). The code increases the begin timestamp with 1 second if it is safe to do so, to prevent the API from returning the same events over and over again. - Some optimisations in handing the processing results (less merges to reduce CPU usage in cases we have lots of events). Testing: - You can test with lots of events using the new mailgun mocking server (Toolbox repo `scripts/mailgun-mock-server`). This can also simulate events that are only returned after x minutes because of storage delays.
2023-02-20 18:44:13 +03:00
/**
* Fetches events that are older than 30 minutes, because then the 'storage' of the Mailgun API is stable. And we are sure we don't miss any events.
* @param {object} options
* @param {number} [options.maxEvents] Not a strict maximum. We stop fetching after we reached the maximum AND received at least one event after begin (not equal) to prevent deadlocks.
*/
async fetchMissing({maxEvents = Infinity} = {}) {
// We start where we left of, or 1,5h ago after a server restart
const begin = this.#fetchMissingData?.lastEventTimestamp ?? this.#fetchMissingData?.lastBegin ?? new Date(Date.now() - TRUST_THRESHOLD_MS * 3);
// Always stop at the time the fetchLatest started fetching on, or maximum 30 minutes ago
const end = new Date(
Math.min(
Date.now() - TRUST_THRESHOLD_MS,
this.#fetchLatestData?.lastBegin?.getTime()
)
);
if (end <= begin) {
Implemented email analytics retrying (#16273) fixes https://github.com/TryGhost/Team/issues/2562 New event fetching loops: - Reworked the analytics fetching algorithm. Instead of starting again where we stopped during the last fetching minus 30 minutes, we now just continue where we stopped. But with ms precision (because no longer database dependent after first fetch), and we stop at NOW - 1 minute to reduce chance of missing events. - Apart from that, a missing fetching loop is introduced. This fetches events that are older than 30 minutes, and just processes all events a second time to make sure we didn't skip any because of storage delays in the Mailgun API. - A new scheduled fetching loop, that allows us to schedule between a given start/end date (currently only persisted in memory, so stops after a reboot) UI and endpoint changes: - New UI to show the state of the analytics 'loops' - New endpoint to request the analytics loop status - New endpoint to schedule analytics - New endpoint to cancel scheduled analytics - Some number formatting improvements, and introduction of 'opened' count in debug screen - Live reload of data in the debug screen Other changes: - This also improves the support for maxEvents. We can now stop a fetching loop after x events without worrying about lost events. This is used to reduce the fetched events in the missing and scheduled event loop (e.g. when the main one is fetching lots of events, we skip the other loops). - Prevents fetching the same events over and over again if no new events come in (because we always started at the same begin timestamp). The code increases the begin timestamp with 1 second if it is safe to do so, to prevent the API from returning the same events over and over again. - Some optimisations in handing the processing results (less merges to reduce CPU usage in cases we have lots of events). Testing: - You can test with lots of events using the new mailgun mocking server (Toolbox repo `scripts/mailgun-mock-server`). This can also simulate events that are only returned after x minutes because of storage delays.
2023-02-20 18:44:13 +03:00
// Skip for now
logging.info('[EmailAnalytics] Skipping fetchMissing because end (' + end + ') is before begin (' + begin + ')');
return 0;
}
Implemented email analytics retrying (#16273) fixes https://github.com/TryGhost/Team/issues/2562 New event fetching loops: - Reworked the analytics fetching algorithm. Instead of starting again where we stopped during the last fetching minus 30 minutes, we now just continue where we stopped. But with ms precision (because no longer database dependent after first fetch), and we stop at NOW - 1 minute to reduce chance of missing events. - Apart from that, a missing fetching loop is introduced. This fetches events that are older than 30 minutes, and just processes all events a second time to make sure we didn't skip any because of storage delays in the Mailgun API. - A new scheduled fetching loop, that allows us to schedule between a given start/end date (currently only persisted in memory, so stops after a reboot) UI and endpoint changes: - New UI to show the state of the analytics 'loops' - New endpoint to request the analytics loop status - New endpoint to schedule analytics - New endpoint to cancel scheduled analytics - Some number formatting improvements, and introduction of 'opened' count in debug screen - Live reload of data in the debug screen Other changes: - This also improves the support for maxEvents. We can now stop a fetching loop after x events without worrying about lost events. This is used to reduce the fetched events in the missing and scheduled event loop (e.g. when the main one is fetching lots of events, we skip the other loops). - Prevents fetching the same events over and over again if no new events come in (because we always started at the same begin timestamp). The code increases the begin timestamp with 1 second if it is safe to do so, to prevent the API from returning the same events over and over again. - Some optimisations in handing the processing results (less merges to reduce CPU usage in cases we have lots of events). Testing: - You can test with lots of events using the new mailgun mocking server (Toolbox repo `scripts/mailgun-mock-server`). This can also simulate events that are only returned after x minutes because of storage delays.
2023-02-20 18:44:13 +03:00
// Create the fetch data object if it doesn't exist yet
if (!this.#fetchMissingData) {
this.#fetchMissingData = {
running: false
};
}
Implemented email analytics retrying (#16273) fixes https://github.com/TryGhost/Team/issues/2562 New event fetching loops: - Reworked the analytics fetching algorithm. Instead of starting again where we stopped during the last fetching minus 30 minutes, we now just continue where we stopped. But with ms precision (because no longer database dependent after first fetch), and we stop at NOW - 1 minute to reduce chance of missing events. - Apart from that, a missing fetching loop is introduced. This fetches events that are older than 30 minutes, and just processes all events a second time to make sure we didn't skip any because of storage delays in the Mailgun API. - A new scheduled fetching loop, that allows us to schedule between a given start/end date (currently only persisted in memory, so stops after a reboot) UI and endpoint changes: - New UI to show the state of the analytics 'loops' - New endpoint to request the analytics loop status - New endpoint to schedule analytics - New endpoint to cancel scheduled analytics - Some number formatting improvements, and introduction of 'opened' count in debug screen - Live reload of data in the debug screen Other changes: - This also improves the support for maxEvents. We can now stop a fetching loop after x events without worrying about lost events. This is used to reduce the fetched events in the missing and scheduled event loop (e.g. when the main one is fetching lots of events, we skip the other loops). - Prevents fetching the same events over and over again if no new events come in (because we always started at the same begin timestamp). The code increases the begin timestamp with 1 second if it is safe to do so, to prevent the API from returning the same events over and over again. - Some optimisations in handing the processing results (less merges to reduce CPU usage in cases we have lots of events). Testing: - You can test with lots of events using the new mailgun mocking server (Toolbox repo `scripts/mailgun-mock-server`). This can also simulate events that are only returned after x minutes because of storage delays.
2023-02-20 18:44:13 +03:00
return await this.#fetchEvents(this.#fetchMissingData, {begin, end, maxEvents});
}
Implemented email analytics retrying (#16273) fixes https://github.com/TryGhost/Team/issues/2562 New event fetching loops: - Reworked the analytics fetching algorithm. Instead of starting again where we stopped during the last fetching minus 30 minutes, we now just continue where we stopped. But with ms precision (because no longer database dependent after first fetch), and we stop at NOW - 1 minute to reduce chance of missing events. - Apart from that, a missing fetching loop is introduced. This fetches events that are older than 30 minutes, and just processes all events a second time to make sure we didn't skip any because of storage delays in the Mailgun API. - A new scheduled fetching loop, that allows us to schedule between a given start/end date (currently only persisted in memory, so stops after a reboot) UI and endpoint changes: - New UI to show the state of the analytics 'loops' - New endpoint to request the analytics loop status - New endpoint to schedule analytics - New endpoint to cancel scheduled analytics - Some number formatting improvements, and introduction of 'opened' count in debug screen - Live reload of data in the debug screen Other changes: - This also improves the support for maxEvents. We can now stop a fetching loop after x events without worrying about lost events. This is used to reduce the fetched events in the missing and scheduled event loop (e.g. when the main one is fetching lots of events, we skip the other loops). - Prevents fetching the same events over and over again if no new events come in (because we always started at the same begin timestamp). The code increases the begin timestamp with 1 second if it is safe to do so, to prevent the API from returning the same events over and over again. - Some optimisations in handing the processing results (less merges to reduce CPU usage in cases we have lots of events). Testing: - You can test with lots of events using the new mailgun mocking server (Toolbox repo `scripts/mailgun-mock-server`). This can also simulate events that are only returned after x minutes because of storage delays.
2023-02-20 18:44:13 +03:00
/**
* Schedule a new fetch that should happen
*/
schedule({begin, end}) {
if (this.#fetchScheduledData && this.#fetchScheduledData.running) {
throw new errors.ValidationError({
message: 'Already fetching scheduled events. Wait for it to finish before scheduling a new one.'
});
}
logging.info('[EmailAnalytics] Scheduling fetch from ' + begin.toISOString() + ' until ' + end.toISOString());
this.#fetchScheduledData = {
running: false,
schedule: {
begin,
end
}
};
}
Implemented email analytics retrying (#16273) fixes https://github.com/TryGhost/Team/issues/2562 New event fetching loops: - Reworked the analytics fetching algorithm. Instead of starting again where we stopped during the last fetching minus 30 minutes, we now just continue where we stopped. But with ms precision (because no longer database dependent after first fetch), and we stop at NOW - 1 minute to reduce chance of missing events. - Apart from that, a missing fetching loop is introduced. This fetches events that are older than 30 minutes, and just processes all events a second time to make sure we didn't skip any because of storage delays in the Mailgun API. - A new scheduled fetching loop, that allows us to schedule between a given start/end date (currently only persisted in memory, so stops after a reboot) UI and endpoint changes: - New UI to show the state of the analytics 'loops' - New endpoint to request the analytics loop status - New endpoint to schedule analytics - New endpoint to cancel scheduled analytics - Some number formatting improvements, and introduction of 'opened' count in debug screen - Live reload of data in the debug screen Other changes: - This also improves the support for maxEvents. We can now stop a fetching loop after x events without worrying about lost events. This is used to reduce the fetched events in the missing and scheduled event loop (e.g. when the main one is fetching lots of events, we skip the other loops). - Prevents fetching the same events over and over again if no new events come in (because we always started at the same begin timestamp). The code increases the begin timestamp with 1 second if it is safe to do so, to prevent the API from returning the same events over and over again. - Some optimisations in handing the processing results (less merges to reduce CPU usage in cases we have lots of events). Testing: - You can test with lots of events using the new mailgun mocking server (Toolbox repo `scripts/mailgun-mock-server`). This can also simulate events that are only returned after x minutes because of storage delays.
2023-02-20 18:44:13 +03:00
cancelScheduled() {
if (this.#fetchScheduledData) {
if (this.#fetchScheduledData.running) {
// Cancel the running fetch
this.#fetchScheduledData.canceled = true;
} else {
this.#fetchScheduledData = null;
}
}
Implemented email analytics retrying (#16273) fixes https://github.com/TryGhost/Team/issues/2562 New event fetching loops: - Reworked the analytics fetching algorithm. Instead of starting again where we stopped during the last fetching minus 30 minutes, we now just continue where we stopped. But with ms precision (because no longer database dependent after first fetch), and we stop at NOW - 1 minute to reduce chance of missing events. - Apart from that, a missing fetching loop is introduced. This fetches events that are older than 30 minutes, and just processes all events a second time to make sure we didn't skip any because of storage delays in the Mailgun API. - A new scheduled fetching loop, that allows us to schedule between a given start/end date (currently only persisted in memory, so stops after a reboot) UI and endpoint changes: - New UI to show the state of the analytics 'loops' - New endpoint to request the analytics loop status - New endpoint to schedule analytics - New endpoint to cancel scheduled analytics - Some number formatting improvements, and introduction of 'opened' count in debug screen - Live reload of data in the debug screen Other changes: - This also improves the support for maxEvents. We can now stop a fetching loop after x events without worrying about lost events. This is used to reduce the fetched events in the missing and scheduled event loop (e.g. when the main one is fetching lots of events, we skip the other loops). - Prevents fetching the same events over and over again if no new events come in (because we always started at the same begin timestamp). The code increases the begin timestamp with 1 second if it is safe to do so, to prevent the API from returning the same events over and over again. - Some optimisations in handing the processing results (less merges to reduce CPU usage in cases we have lots of events). Testing: - You can test with lots of events using the new mailgun mocking server (Toolbox repo `scripts/mailgun-mock-server`). This can also simulate events that are only returned after x minutes because of storage delays.
2023-02-20 18:44:13 +03:00
}
/**
* Continues fetching the scheduled events (does not start one). Resets the scheduled event when received 0 events.
*/
async fetchScheduled({maxEvents = Infinity} = {}) {
if (!this.#fetchScheduledData || !this.#fetchScheduledData.schedule) {
// Nothing scheduled
return 0;
}
Implemented email analytics retrying (#16273) fixes https://github.com/TryGhost/Team/issues/2562 New event fetching loops: - Reworked the analytics fetching algorithm. Instead of starting again where we stopped during the last fetching minus 30 minutes, we now just continue where we stopped. But with ms precision (because no longer database dependent after first fetch), and we stop at NOW - 1 minute to reduce chance of missing events. - Apart from that, a missing fetching loop is introduced. This fetches events that are older than 30 minutes, and just processes all events a second time to make sure we didn't skip any because of storage delays in the Mailgun API. - A new scheduled fetching loop, that allows us to schedule between a given start/end date (currently only persisted in memory, so stops after a reboot) UI and endpoint changes: - New UI to show the state of the analytics 'loops' - New endpoint to request the analytics loop status - New endpoint to schedule analytics - New endpoint to cancel scheduled analytics - Some number formatting improvements, and introduction of 'opened' count in debug screen - Live reload of data in the debug screen Other changes: - This also improves the support for maxEvents. We can now stop a fetching loop after x events without worrying about lost events. This is used to reduce the fetched events in the missing and scheduled event loop (e.g. when the main one is fetching lots of events, we skip the other loops). - Prevents fetching the same events over and over again if no new events come in (because we always started at the same begin timestamp). The code increases the begin timestamp with 1 second if it is safe to do so, to prevent the API from returning the same events over and over again. - Some optimisations in handing the processing results (less merges to reduce CPU usage in cases we have lots of events). Testing: - You can test with lots of events using the new mailgun mocking server (Toolbox repo `scripts/mailgun-mock-server`). This can also simulate events that are only returned after x minutes because of storage delays.
2023-02-20 18:44:13 +03:00
if (this.#fetchScheduledData.canceled) {
// Skip for now
this.#fetchScheduledData = null;
return 0;
}
let begin = this.#fetchScheduledData.schedule.begin;
const end = this.#fetchScheduledData.schedule.end;
if (this.#fetchScheduledData.lastEventTimestamp && this.#fetchScheduledData.lastEventTimestamp > begin) {
// Continue where we left of
begin = this.#fetchScheduledData.lastEventTimestamp;
}
if (end <= begin) {
Implemented email analytics retrying (#16273) fixes https://github.com/TryGhost/Team/issues/2562 New event fetching loops: - Reworked the analytics fetching algorithm. Instead of starting again where we stopped during the last fetching minus 30 minutes, we now just continue where we stopped. But with ms precision (because no longer database dependent after first fetch), and we stop at NOW - 1 minute to reduce chance of missing events. - Apart from that, a missing fetching loop is introduced. This fetches events that are older than 30 minutes, and just processes all events a second time to make sure we didn't skip any because of storage delays in the Mailgun API. - A new scheduled fetching loop, that allows us to schedule between a given start/end date (currently only persisted in memory, so stops after a reboot) UI and endpoint changes: - New UI to show the state of the analytics 'loops' - New endpoint to request the analytics loop status - New endpoint to schedule analytics - New endpoint to cancel scheduled analytics - Some number formatting improvements, and introduction of 'opened' count in debug screen - Live reload of data in the debug screen Other changes: - This also improves the support for maxEvents. We can now stop a fetching loop after x events without worrying about lost events. This is used to reduce the fetched events in the missing and scheduled event loop (e.g. when the main one is fetching lots of events, we skip the other loops). - Prevents fetching the same events over and over again if no new events come in (because we always started at the same begin timestamp). The code increases the begin timestamp with 1 second if it is safe to do so, to prevent the API from returning the same events over and over again. - Some optimisations in handing the processing results (less merges to reduce CPU usage in cases we have lots of events). Testing: - You can test with lots of events using the new mailgun mocking server (Toolbox repo `scripts/mailgun-mock-server`). This can also simulate events that are only returned after x minutes because of storage delays.
2023-02-20 18:44:13 +03:00
// Skip for now
logging.info('[EmailAnalytics] Ending fetchScheduled because end is before begin');
Implemented email analytics retrying (#16273) fixes https://github.com/TryGhost/Team/issues/2562 New event fetching loops: - Reworked the analytics fetching algorithm. Instead of starting again where we stopped during the last fetching minus 30 minutes, we now just continue where we stopped. But with ms precision (because no longer database dependent after first fetch), and we stop at NOW - 1 minute to reduce chance of missing events. - Apart from that, a missing fetching loop is introduced. This fetches events that are older than 30 minutes, and just processes all events a second time to make sure we didn't skip any because of storage delays in the Mailgun API. - A new scheduled fetching loop, that allows us to schedule between a given start/end date (currently only persisted in memory, so stops after a reboot) UI and endpoint changes: - New UI to show the state of the analytics 'loops' - New endpoint to request the analytics loop status - New endpoint to schedule analytics - New endpoint to cancel scheduled analytics - Some number formatting improvements, and introduction of 'opened' count in debug screen - Live reload of data in the debug screen Other changes: - This also improves the support for maxEvents. We can now stop a fetching loop after x events without worrying about lost events. This is used to reduce the fetched events in the missing and scheduled event loop (e.g. when the main one is fetching lots of events, we skip the other loops). - Prevents fetching the same events over and over again if no new events come in (because we always started at the same begin timestamp). The code increases the begin timestamp with 1 second if it is safe to do so, to prevent the API from returning the same events over and over again. - Some optimisations in handing the processing results (less merges to reduce CPU usage in cases we have lots of events). Testing: - You can test with lots of events using the new mailgun mocking server (Toolbox repo `scripts/mailgun-mock-server`). This can also simulate events that are only returned after x minutes because of storage delays.
2023-02-20 18:44:13 +03:00
this.#fetchScheduledData = null;
return 0;
}
const count = await this.#fetchEvents(this.#fetchScheduledData, {begin, end, maxEvents});
if (count === 0 || this.#fetchScheduledData.canceled) {
// Reset the scheduled fetch
this.#fetchScheduledData = null;
}
return count;
}
Implemented email analytics retrying (#16273) fixes https://github.com/TryGhost/Team/issues/2562 New event fetching loops: - Reworked the analytics fetching algorithm. Instead of starting again where we stopped during the last fetching minus 30 minutes, we now just continue where we stopped. But with ms precision (because no longer database dependent after first fetch), and we stop at NOW - 1 minute to reduce chance of missing events. - Apart from that, a missing fetching loop is introduced. This fetches events that are older than 30 minutes, and just processes all events a second time to make sure we didn't skip any because of storage delays in the Mailgun API. - A new scheduled fetching loop, that allows us to schedule between a given start/end date (currently only persisted in memory, so stops after a reboot) UI and endpoint changes: - New UI to show the state of the analytics 'loops' - New endpoint to request the analytics loop status - New endpoint to schedule analytics - New endpoint to cancel scheduled analytics - Some number formatting improvements, and introduction of 'opened' count in debug screen - Live reload of data in the debug screen Other changes: - This also improves the support for maxEvents. We can now stop a fetching loop after x events without worrying about lost events. This is used to reduce the fetched events in the missing and scheduled event loop (e.g. when the main one is fetching lots of events, we skip the other loops). - Prevents fetching the same events over and over again if no new events come in (because we always started at the same begin timestamp). The code increases the begin timestamp with 1 second if it is safe to do so, to prevent the API from returning the same events over and over again. - Some optimisations in handing the processing results (less merges to reduce CPU usage in cases we have lots of events). Testing: - You can test with lots of events using the new mailgun mocking server (Toolbox repo `scripts/mailgun-mock-server`). This can also simulate events that are only returned after x minutes because of storage delays.
2023-02-20 18:44:13 +03:00
/**
* Start fetching analytics and store the data of the progress inside fetchData
* @param {FetchData} fetchData
* @param {object} options
* @param {Date} options.begin
* @param {Date} options.end
* @param {number} [options.maxEvents] Not a strict maximum. We stop fetching after we reached the maximum AND received at least one event after begin (not equal) to prevent deadlocks.
*/
async #fetchEvents(fetchData, {begin, end, maxEvents = Infinity}) {
// Start where we left of, or the last stored event in the database, or start 30 minutes ago if we have nothing available
logging.info('[EmailAnalytics] Fetching from ' + begin.toISOString() + ' until ' + end.toISOString() + ' (maxEvents: ' + maxEvents + ')');
// Store that we started fetching
fetchData.running = true;
fetchData.lastStarted = new Date();
fetchData.lastBegin = begin;
let lastAggregation = Date.now();
let eventCount = 0;
// We keep the processing result here, so we also have a result in case of failures
let processingResult = new EventProcessingResult();
let error = null;
const processBatch = async (events) => {
// Even if the fetching is interrupted because of an error, we still store the last event timestamp
await this.processEventBatch(events, processingResult, fetchData);
eventCount += events.length;
// Every 5 minutes or 5000 members we do an aggregation and clear the processingResult
// Otherwise we need to loop a lot of members afterwards, and this takes too long without updating the stat counts in between
if (Date.now() - lastAggregation > 5 * 60 * 1000 || processingResult.memberIds.length > 5000) {
// Aggregate and clear the processingResult
// We do this here because otherwise it could take a long time before the new events are visible in the stats
try {
await this.aggregateStats(processingResult);
lastAggregation = Date.now();
processingResult = new EventProcessingResult();
} catch (err) {
logging.error('[EmailAnalytics] Error while aggregating stats');
logging.error(err);
}
}
if (fetchData.canceled) {
throw new errors.InternalServerError({
message: 'Fetching canceled'
});
}
};
try {
for (const provider of this.providers) {
await provider.fetchLatest(processBatch, {begin, end, maxEvents});
}
2023-02-27 17:00:18 +03:00
logging.info('[EmailAnalytics] Fetching finished');
Implemented email analytics retrying (#16273) fixes https://github.com/TryGhost/Team/issues/2562 New event fetching loops: - Reworked the analytics fetching algorithm. Instead of starting again where we stopped during the last fetching minus 30 minutes, we now just continue where we stopped. But with ms precision (because no longer database dependent after first fetch), and we stop at NOW - 1 minute to reduce chance of missing events. - Apart from that, a missing fetching loop is introduced. This fetches events that are older than 30 minutes, and just processes all events a second time to make sure we didn't skip any because of storage delays in the Mailgun API. - A new scheduled fetching loop, that allows us to schedule between a given start/end date (currently only persisted in memory, so stops after a reboot) UI and endpoint changes: - New UI to show the state of the analytics 'loops' - New endpoint to request the analytics loop status - New endpoint to schedule analytics - New endpoint to cancel scheduled analytics - Some number formatting improvements, and introduction of 'opened' count in debug screen - Live reload of data in the debug screen Other changes: - This also improves the support for maxEvents. We can now stop a fetching loop after x events without worrying about lost events. This is used to reduce the fetched events in the missing and scheduled event loop (e.g. when the main one is fetching lots of events, we skip the other loops). - Prevents fetching the same events over and over again if no new events come in (because we always started at the same begin timestamp). The code increases the begin timestamp with 1 second if it is safe to do so, to prevent the API from returning the same events over and over again. - Some optimisations in handing the processing results (less merges to reduce CPU usage in cases we have lots of events). Testing: - You can test with lots of events using the new mailgun mocking server (Toolbox repo `scripts/mailgun-mock-server`). This can also simulate events that are only returned after x minutes because of storage delays.
2023-02-20 18:44:13 +03:00
} catch (err) {
if (err.message !== 'Fetching canceled') {
logging.error('[EmailAnalytics] Error while fetching');
logging.error(err);
error = err;
} else {
logging.error('[EmailAnalytics] Canceled fetching');
}
}
// Aggregate
try {
await this.aggregateStats(processingResult);
} catch (err) {
logging.error('[EmailAnalytics] Error while aggregating stats');
logging.error(err);
if (!error) {
error = err;
}
}
// Small trick: if reached the end of new events, we are going to keep
// fetching the same events because 'begin' won't change
// So if we didn't have errors while fetching, and total events < maxEvents, increase lastEventTimestamp with one second
if (!error && eventCount > 0 && eventCount < maxEvents && fetchData.lastEventTimestamp && fetchData.lastEventTimestamp.getTime() < Date.now() - 2000) {
logging.info('[EmailAnalytics] Reached end of new events, increasing lastEventTimestamp with one second');
fetchData.lastEventTimestamp = new Date(fetchData.lastEventTimestamp.getTime() + 1000);
}
Implemented email analytics retrying (#16273) fixes https://github.com/TryGhost/Team/issues/2562 New event fetching loops: - Reworked the analytics fetching algorithm. Instead of starting again where we stopped during the last fetching minus 30 minutes, we now just continue where we stopped. But with ms precision (because no longer database dependent after first fetch), and we stop at NOW - 1 minute to reduce chance of missing events. - Apart from that, a missing fetching loop is introduced. This fetches events that are older than 30 minutes, and just processes all events a second time to make sure we didn't skip any because of storage delays in the Mailgun API. - A new scheduled fetching loop, that allows us to schedule between a given start/end date (currently only persisted in memory, so stops after a reboot) UI and endpoint changes: - New UI to show the state of the analytics 'loops' - New endpoint to request the analytics loop status - New endpoint to schedule analytics - New endpoint to cancel scheduled analytics - Some number formatting improvements, and introduction of 'opened' count in debug screen - Live reload of data in the debug screen Other changes: - This also improves the support for maxEvents. We can now stop a fetching loop after x events without worrying about lost events. This is used to reduce the fetched events in the missing and scheduled event loop (e.g. when the main one is fetching lots of events, we skip the other loops). - Prevents fetching the same events over and over again if no new events come in (because we always started at the same begin timestamp). The code increases the begin timestamp with 1 second if it is safe to do so, to prevent the API from returning the same events over and over again. - Some optimisations in handing the processing results (less merges to reduce CPU usage in cases we have lots of events). Testing: - You can test with lots of events using the new mailgun mocking server (Toolbox repo `scripts/mailgun-mock-server`). This can also simulate events that are only returned after x minutes because of storage delays.
2023-02-20 18:44:13 +03:00
fetchData.running = false;
if (error) {
throw error;
}
return eventCount;
}
/**
* @param {any[]} events
* @param {FetchData} fetchData
*/
async processEventBatch(events, result, fetchData) {
const processStart = Date.now();
for (const event of events) {
Added new email event processor (#15879) fixes https://github.com/TryGhost/Team/issues/2310 This moves the processing of the events from the event-processor to a new email-event-processor in the email-service package. - The `EmailEventProcessor` only translates events from providerId/emailId to their known emailId, memberId and recipientId, and dispatches the corresponding events. - Since `EmailEventProcessor` runs in a separate worker thread, we can't listen for the dispatched events on the main thread. To accomplish this communication, the events dispatched from the `EmailEventProcessor` class are 'posted' via the postMessage method and redispatched on the main thread. - A new `EmailEventStorage` class reacts to the email events and stores it in the database. This code mostly corresponds to the (now deleted) subclass of the old `EmailEventProcessor` - Updating a members last_seen_at timestamp has moved to the lastSeenAtUpdater. - Email events no longer store `ObjectID` because these are not encodable across threads via postMessage - Includes new E2E tests that test the storage of all supported Mailgun events. Note that in these tests we run the processing on the main thread instead of on a separate thread (couldn't do this because stubbing is not possible across threads) There are some missing pieces that will get added in later PRs (this PR focuses on porting the existing functionality): - Handling temporary failures/bounces - Capturing the error messages of bounce events
2022-11-29 13:15:19 +03:00
const batchResult = await this.processEvent(event);
Implemented email analytics retrying (#16273) fixes https://github.com/TryGhost/Team/issues/2562 New event fetching loops: - Reworked the analytics fetching algorithm. Instead of starting again where we stopped during the last fetching minus 30 minutes, we now just continue where we stopped. But with ms precision (because no longer database dependent after first fetch), and we stop at NOW - 1 minute to reduce chance of missing events. - Apart from that, a missing fetching loop is introduced. This fetches events that are older than 30 minutes, and just processes all events a second time to make sure we didn't skip any because of storage delays in the Mailgun API. - A new scheduled fetching loop, that allows us to schedule between a given start/end date (currently only persisted in memory, so stops after a reboot) UI and endpoint changes: - New UI to show the state of the analytics 'loops' - New endpoint to request the analytics loop status - New endpoint to schedule analytics - New endpoint to cancel scheduled analytics - Some number formatting improvements, and introduction of 'opened' count in debug screen - Live reload of data in the debug screen Other changes: - This also improves the support for maxEvents. We can now stop a fetching loop after x events without worrying about lost events. This is used to reduce the fetched events in the missing and scheduled event loop (e.g. when the main one is fetching lots of events, we skip the other loops). - Prevents fetching the same events over and over again if no new events come in (because we always started at the same begin timestamp). The code increases the begin timestamp with 1 second if it is safe to do so, to prevent the API from returning the same events over and over again. - Some optimisations in handing the processing results (less merges to reduce CPU usage in cases we have lots of events). Testing: - You can test with lots of events using the new mailgun mocking server (Toolbox repo `scripts/mailgun-mock-server`). This can also simulate events that are only returned after x minutes because of storage delays.
2023-02-20 18:44:13 +03:00
// Save last event timestamp
if (!fetchData.lastEventTimestamp || (event.timestamp && event.timestamp > fetchData.lastEventTimestamp)) {
fetchData.lastEventTimestamp = event.timestamp;
}
result.merge(batchResult);
}
Implemented email analytics retrying (#16273) fixes https://github.com/TryGhost/Team/issues/2562 New event fetching loops: - Reworked the analytics fetching algorithm. Instead of starting again where we stopped during the last fetching minus 30 minutes, we now just continue where we stopped. But with ms precision (because no longer database dependent after first fetch), and we stop at NOW - 1 minute to reduce chance of missing events. - Apart from that, a missing fetching loop is introduced. This fetches events that are older than 30 minutes, and just processes all events a second time to make sure we didn't skip any because of storage delays in the Mailgun API. - A new scheduled fetching loop, that allows us to schedule between a given start/end date (currently only persisted in memory, so stops after a reboot) UI and endpoint changes: - New UI to show the state of the analytics 'loops' - New endpoint to request the analytics loop status - New endpoint to schedule analytics - New endpoint to cancel scheduled analytics - Some number formatting improvements, and introduction of 'opened' count in debug screen - Live reload of data in the debug screen Other changes: - This also improves the support for maxEvents. We can now stop a fetching loop after x events without worrying about lost events. This is used to reduce the fetched events in the missing and scheduled event loop (e.g. when the main one is fetching lots of events, we skip the other loops). - Prevents fetching the same events over and over again if no new events come in (because we always started at the same begin timestamp). The code increases the begin timestamp with 1 second if it is safe to do so, to prevent the API from returning the same events over and over again. - Some optimisations in handing the processing results (less merges to reduce CPU usage in cases we have lots of events). Testing: - You can test with lots of events using the new mailgun mocking server (Toolbox repo `scripts/mailgun-mock-server`). This can also simulate events that are only returned after x minutes because of storage delays.
2023-02-20 18:44:13 +03:00
const processEnd = Date.now();
const time = processEnd - processStart;
if (time > 1000) {
// This is a means to show in the logs that the analytics job is still alive.
logging.warn(`[EmailAnalytics] Processing event batch took ${(time / 1000).toFixed(1)}s`);
}
}
Added new email event processor (#15879) fixes https://github.com/TryGhost/Team/issues/2310 This moves the processing of the events from the event-processor to a new email-event-processor in the email-service package. - The `EmailEventProcessor` only translates events from providerId/emailId to their known emailId, memberId and recipientId, and dispatches the corresponding events. - Since `EmailEventProcessor` runs in a separate worker thread, we can't listen for the dispatched events on the main thread. To accomplish this communication, the events dispatched from the `EmailEventProcessor` class are 'posted' via the postMessage method and redispatched on the main thread. - A new `EmailEventStorage` class reacts to the email events and stores it in the database. This code mostly corresponds to the (now deleted) subclass of the old `EmailEventProcessor` - Updating a members last_seen_at timestamp has moved to the lastSeenAtUpdater. - Email events no longer store `ObjectID` because these are not encodable across threads via postMessage - Includes new E2E tests that test the storage of all supported Mailgun events. Note that in these tests we run the processing on the main thread instead of on a separate thread (couldn't do this because stubbing is not possible across threads) There are some missing pieces that will get added in later PRs (this PR focuses on porting the existing functionality): - Handling temporary failures/bounces - Capturing the error messages of bounce events
2022-11-29 13:15:19 +03:00
/**
*
Implemented email analytics retrying (#16273) fixes https://github.com/TryGhost/Team/issues/2562 New event fetching loops: - Reworked the analytics fetching algorithm. Instead of starting again where we stopped during the last fetching minus 30 minutes, we now just continue where we stopped. But with ms precision (because no longer database dependent after first fetch), and we stop at NOW - 1 minute to reduce chance of missing events. - Apart from that, a missing fetching loop is introduced. This fetches events that are older than 30 minutes, and just processes all events a second time to make sure we didn't skip any because of storage delays in the Mailgun API. - A new scheduled fetching loop, that allows us to schedule between a given start/end date (currently only persisted in memory, so stops after a reboot) UI and endpoint changes: - New UI to show the state of the analytics 'loops' - New endpoint to request the analytics loop status - New endpoint to schedule analytics - New endpoint to cancel scheduled analytics - Some number formatting improvements, and introduction of 'opened' count in debug screen - Live reload of data in the debug screen Other changes: - This also improves the support for maxEvents. We can now stop a fetching loop after x events without worrying about lost events. This is used to reduce the fetched events in the missing and scheduled event loop (e.g. when the main one is fetching lots of events, we skip the other loops). - Prevents fetching the same events over and over again if no new events come in (because we always started at the same begin timestamp). The code increases the begin timestamp with 1 second if it is safe to do so, to prevent the API from returning the same events over and over again. - Some optimisations in handing the processing results (less merges to reduce CPU usage in cases we have lots of events). Testing: - You can test with lots of events using the new mailgun mocking server (Toolbox repo `scripts/mailgun-mock-server`). This can also simulate events that are only returned after x minutes because of storage delays.
2023-02-20 18:44:13 +03:00
* @param {{id: string, type: any; severity: any; recipientEmail: any; emailId?: string; providerId: string; timestamp: Date; error: {code: number; message: string; enhandedCode: string|number} | null}} event
Added new email event processor (#15879) fixes https://github.com/TryGhost/Team/issues/2310 This moves the processing of the events from the event-processor to a new email-event-processor in the email-service package. - The `EmailEventProcessor` only translates events from providerId/emailId to their known emailId, memberId and recipientId, and dispatches the corresponding events. - Since `EmailEventProcessor` runs in a separate worker thread, we can't listen for the dispatched events on the main thread. To accomplish this communication, the events dispatched from the `EmailEventProcessor` class are 'posted' via the postMessage method and redispatched on the main thread. - A new `EmailEventStorage` class reacts to the email events and stores it in the database. This code mostly corresponds to the (now deleted) subclass of the old `EmailEventProcessor` - Updating a members last_seen_at timestamp has moved to the lastSeenAtUpdater. - Email events no longer store `ObjectID` because these are not encodable across threads via postMessage - Includes new E2E tests that test the storage of all supported Mailgun events. Note that in these tests we run the processing on the main thread instead of on a separate thread (couldn't do this because stubbing is not possible across threads) There are some missing pieces that will get added in later PRs (this PR focuses on porting the existing functionality): - Handling temporary failures/bounces - Capturing the error messages of bounce events
2022-11-29 13:15:19 +03:00
* @returns {Promise<EventProcessingResult>}
*/
async processEvent(event) {
if (event.type === 'delivered') {
const recipient = await this.eventProcessor.handleDelivered({emailId: event.emailId, providerId: event.providerId, email: event.recipientEmail}, event.timestamp);
if (recipient) {
return new EventProcessingResult({
delivered: 1,
emailIds: [recipient.emailId],
memberIds: [recipient.memberId]
});
}
return new EventProcessingResult({unprocessable: 1});
}
if (event.type === 'opened') {
const recipient = await this.eventProcessor.handleOpened({emailId: event.emailId, providerId: event.providerId, email: event.recipientEmail}, event.timestamp);
if (recipient) {
return new EventProcessingResult({
opened: 1,
emailIds: [recipient.emailId],
memberIds: [recipient.memberId]
});
}
return new EventProcessingResult({unprocessable: 1});
}
if (event.type === 'failed') {
if (event.severity === 'permanent') {
const recipient = await this.eventProcessor.handlePermanentFailed({emailId: event.emailId, providerId: event.providerId, email: event.recipientEmail}, {id: event.id, timestamp: event.timestamp, error: event.error});
Added new email event processor (#15879) fixes https://github.com/TryGhost/Team/issues/2310 This moves the processing of the events from the event-processor to a new email-event-processor in the email-service package. - The `EmailEventProcessor` only translates events from providerId/emailId to their known emailId, memberId and recipientId, and dispatches the corresponding events. - Since `EmailEventProcessor` runs in a separate worker thread, we can't listen for the dispatched events on the main thread. To accomplish this communication, the events dispatched from the `EmailEventProcessor` class are 'posted' via the postMessage method and redispatched on the main thread. - A new `EmailEventStorage` class reacts to the email events and stores it in the database. This code mostly corresponds to the (now deleted) subclass of the old `EmailEventProcessor` - Updating a members last_seen_at timestamp has moved to the lastSeenAtUpdater. - Email events no longer store `ObjectID` because these are not encodable across threads via postMessage - Includes new E2E tests that test the storage of all supported Mailgun events. Note that in these tests we run the processing on the main thread instead of on a separate thread (couldn't do this because stubbing is not possible across threads) There are some missing pieces that will get added in later PRs (this PR focuses on porting the existing functionality): - Handling temporary failures/bounces - Capturing the error messages of bounce events
2022-11-29 13:15:19 +03:00
if (recipient) {
return new EventProcessingResult({
permanentFailed: 1,
emailIds: [recipient.emailId],
memberIds: [recipient.memberId]
});
}
return new EventProcessingResult({unprocessable: 1});
} else {
const recipient = await this.eventProcessor.handleTemporaryFailed({emailId: event.emailId, providerId: event.providerId, email: event.recipientEmail}, {id: event.id, timestamp: event.timestamp, error: event.error});
Added new email event processor (#15879) fixes https://github.com/TryGhost/Team/issues/2310 This moves the processing of the events from the event-processor to a new email-event-processor in the email-service package. - The `EmailEventProcessor` only translates events from providerId/emailId to their known emailId, memberId and recipientId, and dispatches the corresponding events. - Since `EmailEventProcessor` runs in a separate worker thread, we can't listen for the dispatched events on the main thread. To accomplish this communication, the events dispatched from the `EmailEventProcessor` class are 'posted' via the postMessage method and redispatched on the main thread. - A new `EmailEventStorage` class reacts to the email events and stores it in the database. This code mostly corresponds to the (now deleted) subclass of the old `EmailEventProcessor` - Updating a members last_seen_at timestamp has moved to the lastSeenAtUpdater. - Email events no longer store `ObjectID` because these are not encodable across threads via postMessage - Includes new E2E tests that test the storage of all supported Mailgun events. Note that in these tests we run the processing on the main thread instead of on a separate thread (couldn't do this because stubbing is not possible across threads) There are some missing pieces that will get added in later PRs (this PR focuses on porting the existing functionality): - Handling temporary failures/bounces - Capturing the error messages of bounce events
2022-11-29 13:15:19 +03:00
if (recipient) {
return new EventProcessingResult({
temporaryFailed: 1,
emailIds: [recipient.emailId],
memberIds: [recipient.memberId]
});
}
return new EventProcessingResult({unprocessable: 1});
}
}
if (event.type === 'unsubscribed') {
const recipient = await this.eventProcessor.handleUnsubscribed({emailId: event.emailId, providerId: event.providerId, email: event.recipientEmail}, event.timestamp);
if (recipient) {
return new EventProcessingResult({
unsubscribed: 1,
emailIds: [recipient.emailId],
memberIds: [recipient.memberId]
});
}
return new EventProcessingResult({unprocessable: 1});
}
if (event.type === 'complained') {
const recipient = await this.eventProcessor.handleComplained({emailId: event.emailId, providerId: event.providerId, email: event.recipientEmail}, event.timestamp);
if (recipient) {
return new EventProcessingResult({
complained: 1,
emailIds: [recipient.emailId],
memberIds: [recipient.memberId]
});
}
return new EventProcessingResult({unprocessable: 1});
}
return new EventProcessingResult({unhandled: 1});
}
async aggregateStats({emailIds = [], memberIds = []}) {
Implemented email analytics retrying (#16273) fixes https://github.com/TryGhost/Team/issues/2562 New event fetching loops: - Reworked the analytics fetching algorithm. Instead of starting again where we stopped during the last fetching minus 30 minutes, we now just continue where we stopped. But with ms precision (because no longer database dependent after first fetch), and we stop at NOW - 1 minute to reduce chance of missing events. - Apart from that, a missing fetching loop is introduced. This fetches events that are older than 30 minutes, and just processes all events a second time to make sure we didn't skip any because of storage delays in the Mailgun API. - A new scheduled fetching loop, that allows us to schedule between a given start/end date (currently only persisted in memory, so stops after a reboot) UI and endpoint changes: - New UI to show the state of the analytics 'loops' - New endpoint to request the analytics loop status - New endpoint to schedule analytics - New endpoint to cancel scheduled analytics - Some number formatting improvements, and introduction of 'opened' count in debug screen - Live reload of data in the debug screen Other changes: - This also improves the support for maxEvents. We can now stop a fetching loop after x events without worrying about lost events. This is used to reduce the fetched events in the missing and scheduled event loop (e.g. when the main one is fetching lots of events, we skip the other loops). - Prevents fetching the same events over and over again if no new events come in (because we always started at the same begin timestamp). The code increases the begin timestamp with 1 second if it is safe to do so, to prevent the API from returning the same events over and over again. - Some optimisations in handing the processing results (less merges to reduce CPU usage in cases we have lots of events). Testing: - You can test with lots of events using the new mailgun mocking server (Toolbox repo `scripts/mailgun-mock-server`). This can also simulate events that are only returned after x minutes because of storage delays.
2023-02-20 18:44:13 +03:00
logging.info(`[EmailAnalytics] Aggregating for ${emailIds.length} emails`);
for (const emailId of emailIds) {
await this.aggregateEmailStats(emailId);
}
Implemented email analytics retrying (#16273) fixes https://github.com/TryGhost/Team/issues/2562 New event fetching loops: - Reworked the analytics fetching algorithm. Instead of starting again where we stopped during the last fetching minus 30 minutes, we now just continue where we stopped. But with ms precision (because no longer database dependent after first fetch), and we stop at NOW - 1 minute to reduce chance of missing events. - Apart from that, a missing fetching loop is introduced. This fetches events that are older than 30 minutes, and just processes all events a second time to make sure we didn't skip any because of storage delays in the Mailgun API. - A new scheduled fetching loop, that allows us to schedule between a given start/end date (currently only persisted in memory, so stops after a reboot) UI and endpoint changes: - New UI to show the state of the analytics 'loops' - New endpoint to request the analytics loop status - New endpoint to schedule analytics - New endpoint to cancel scheduled analytics - Some number formatting improvements, and introduction of 'opened' count in debug screen - Live reload of data in the debug screen Other changes: - This also improves the support for maxEvents. We can now stop a fetching loop after x events without worrying about lost events. This is used to reduce the fetched events in the missing and scheduled event loop (e.g. when the main one is fetching lots of events, we skip the other loops). - Prevents fetching the same events over and over again if no new events come in (because we always started at the same begin timestamp). The code increases the begin timestamp with 1 second if it is safe to do so, to prevent the API from returning the same events over and over again. - Some optimisations in handing the processing results (less merges to reduce CPU usage in cases we have lots of events). Testing: - You can test with lots of events using the new mailgun mocking server (Toolbox repo `scripts/mailgun-mock-server`). This can also simulate events that are only returned after x minutes because of storage delays.
2023-02-20 18:44:13 +03:00
logging.info(`[EmailAnalytics] Aggregating for ${memberIds.length} members`);
for (const memberId of memberIds) {
await this.aggregateMemberStats(memberId);
}
}
async aggregateEmailStats(emailId) {
return this.queries.aggregateEmailStats(emailId);
}
async aggregateMemberStats(memberId) {
return this.queries.aggregateMemberStats(memberId);
}
};