ref
https://linear.app/tryghost/issue/ENG-902/add-an-optional-timeout-in-the-redis-cache-adapter-in-case-redis
- Added an optional timeout parameter to AdapterCacheRedis, so that the
`get(key)` method will return `null` after the timeout if it hasn't
received a response from Redis
- When load testing the `LinkRedirectRepository` with the Redis cache
enabled on staging, we noticed that for some reason Redis stopped
responding to commands for ~30 seconds.
- The `LinkRedirectRepository` was waiting for the Redis cache to
respond and resulted in a drastic increase in response times for link
redirects
- This change will allow us to set a timeout on the `get(key)` method,
so that if Redis doesn't respond within the timeout, the method will
return `null` as if it were a cache miss.
- Then the `LinkRedirectRepository` will fall back to the database and
return the link redirect from the database instead of waiting
indefinitely for Redis to respond
This updates the AdapterCacheRedis instance to be able to handle updating
itself when reading from the cache. For this to work we need to pass a
`fetchData` function to the `get` method.
In the case of a cache miss, we will read the data via the `fetchData`
function, and store it in the cache, before returning the value to the caller.
When coupled with a `refreshAheadFactor` config, we will go a step further and
implement the "Refresh Ahead" caching strategy. What this means is that we will
refresh the contents of the cache in the background, this happens on a cache
read and only once the data in the cache has only a certain percentage of the
TTL left, which is set as a decimal value between 0 and 1.
e.g.
ttl = 100s
refreshAheadFactor = 0.2;
Any read from the cache that happens _after_ 80s will do a background refresh
The main changes are:
- Updating the pipeline to allow for doing a background refresh of the
cache
- Remove the use of the EventAwareCacheWrapper for the posts public
cache
### Background refresh
This is just an initial implementation, and tbh it doesn't sit right
with me that the logic for this is in the pipeline - I think this should
sit in the cache implementation itself, and then we call out to it with
something like: `cache.get(key, fetchData)` and then the updates can
happen internally.
The `cache-manager` project actually has a method like this called
`wrap` - but every time I've used it it hangs, and debugging was a pain,
so I don't really trust it.
### EventAwareCacheWrapper
This is such a small amount of logic, I don't think it's worth creating
an entire wrapper for it, at least not a class based one. I would be
happy to refactor this to use a `Proxy` too, so that we don't have to
add methods to it each time we wanna change the underlying cache
implementation.
no issue
Began tracking the following metrics when purging the redis cache:
1. cache_reset_scan - total time to scan the keyspace
2. cache_reset_delete - total time to delete all the matching keys
3. cache_reset - total time in ms to reset the cache
4. cache_reset_keys - total number of keys deleted
We can reduce the granularity of these alerts to reduce the load on
elastic eventually, but for now it would be nice to collect metrics at
this granularity so we can optimize the cache purging performance.
no refs
- This enables supplying a username to connect to redis with when connection to an individual node rather than a cluster.
- It also allows for extra options to be given to the ioredis store used in the background.
closes https://github.com/TryGhost/Arch/issues/85
- Added a cache configuration option to signal "reuse of redis connection" for Redis cache adapter. The connection reuse it turned on by default to be shared between caches. They rely on unique "keyPrefix" structure, so there is no collision side-effects when reusing same Redis Store.
- The Redis connection options like "ttl" are shared with the first connection that's crated. So if there's a need to have unique configuration, a separate connection has to be created by passing `"reuseConnection": false` parameter
Previously, the adapter was only built for a redis cluster connection. Meaning
if you tried to use if with a redis that was a single node it would fail as it
tries to find a primary node. In a single node setup there is no primary node,
just the one main node. So, this update tricks the adapter into thinking it has
found a pimary node by returning the whole connection (to the single node) when
the constructor is note a cluster.
As discussed with the product team we want to enforce kebab-case file names for
all files, with the exception of files which export a single class, in which
case they should be PascalCase and reflect the class which they export.
This will help find classes faster, and should push better naming for them too.
Some files and packages have been excluded from this linting, specifically when
a library or framework depends on the naming of a file for the functionality
e.g. Ember, knex-migrator, adapter-manager
refs https://github.com/TryGhost/Toolbox/issues/520
- The cluster config is taking over local adapter ttl configuration - the priority should be reverse adapter config first followed by cluster config
- In addition if we add nested config merging to adapter manager we could achieve the same effect by having per-adapter configuraiton with "clusterConfig.options.ttl" value
refs https://github.com/TryGhost/Toolbox/issues/515
- This implementation allows to use Redis cluster as a caching adapter. The cache adapter can be configured through same adapter configuration interface as others. It accepts following config values:
- "ttl" - time in SECONDS for stored entity to live in Redis cache
- "keyPrefix" - special cache key prefix to use with stored entities
- "host" / "port" / "password" / "clusterConfig" - Redis instance specific configs
- Set test coverage to non-standard 75% because the code mostly consists of the glue code that would require unnecessary Redis server mocking and would be a bad ROI. This module has been used in production for a long time already, so can be considered pretty stable.
refs https://github.com/TryGhost/Toolbox/issues/515
- Redis-based caches can be used on hosted-environments to store information with high memory impact - when in-memory caches would be too impractical to use
- This is a placeholder package for a cluster-aware Redis cache implementation compatible with Ghost's cache adapter interface (a41d351f16/packages/adapter-base-cache)
