redis-cache:metrics key growing up infinitely

[kamilgrzegorczyk]

redis-cache:metrics key is growing over time

Description

The issue happens only on one of dozens of sites I have redis installed on. It started happening over a year ago.
For whatever reason metrics key grows over the few days and gaining even 400-500megabytes.

The site has basic redis configuration

Expected Behavior

Metrics consume max 5-10 megabytes of memory

Actual Behavior

Metrics key consume 500 megabytes (see attached screen)

Possible Fix

Disabling metrics for that particular site (dummy fix)

Steps to Reproduce

I can't provide any steps for reproducing that bug

Additional context

It just consumes a lot of memory which instead should be available for the "real data" keys

The site is running on two domains (completely separated, shared codebase but separate infra) and the issue exists on both.
Both of the domains are med size traffic (tens of thousands users per day)

Environment

As the bug occurs since over a year therefore it was happening under a range of versions up until the most recent
Plugin version - most probably since 2.3.x
WP version - most probably at least since 6.2.x
PHP version 8.0.x-8.1.x

Site is converted to a multisite since 2 months but the issue was happening for over a year when it was still a single site.

[kamilgrzegorczyk]

Forgotten screenshot

[tillkruss]

Please post your diagnostics.

[kamilgrzegorczyk]

Here you are

Status: Connected
Client: PhpRedis (v6.0.2)
Drop-in: Valid
Disabled: No
Ping: 1
Errors: []
PhpRedis: 6.0.2
Relay: Not loaded
Predis: 2.1.2
Credis: Not loaded
PHP Version: 8.1.26
Plugin Version: 2.5.0
Redis Version: 7.0.7
Multisite: Yes
Metrics: Enabled
Metrics recorded: 982398
Filesystem: File modifications are not allowed.
Global Prefix: ""
Blog Prefix: 1
WP_REDIS_CLIENT: "phpredis"
WP_REDIS_HOST: "redacted-out"
WP_REDIS_PORT: 6379
WP_REDIS_DATABASE: 1
WP_REDIS_TIMEOUT: 1
WP_REDIS_READ_TIMEOUT: 1
WP_REDIS_PREFIX: "bnd_"
WP_REDIS_PLUGIN_PATH: "/var/redacted-out/"
Global Groups: [
    "blog-details",
    "blog-id-cache",
    "blog-lookup",
    "global-posts",
    "networks",
    "rss",
    "sites",
    "site-details",
    "site-lookup",
    "site-options",
    "site-transient",
    "users",
    "useremail",
    "userlogins",
    "usermeta",
    "user_meta",
    "userslugs",
    "redis-cache",
    "blog_meta",
    "network-queries",
    "site-queries",
    "user-queries"
]
Ignored Groups: [
    "counts",
    "plugins",
    "theme_json",
    "themes"
]
Unflushable Groups: []
Groups Types: {
    "blog-details": "global",
    "blog-id-cache": "global",
    "blog-lookup": "global",
    "global-posts": "global",
    "networks": "global",
    "rss": "global",
    "sites": "global",
    "site-details": "global",
    "site-lookup": "global",
    "site-options": "global",
    "site-transient": "global",
    "users": "global",
    "useremail": "global",
    "userlogins": "global",
    "usermeta": "global",
    "user_meta": "global",
    "userslugs": "global",
    "redis-cache": "global",
    "blog_meta": "global",
    "network-queries": "global",
    "site-queries": "global",
    "user-queries": "global",
    "counts": "ignored",
    "plugins": "ignored",
    "theme_json": "ignored",
    "themes": "ignored"
}
Drop-ins: [
    "Redis Object Cache Drop-In v2.5.0 by Till Krüss"
]

[tillkruss]

The config looks good. You have 982,398 metrics recorded. If it's taking up too much space, use the WP_REDIS_DISABLE_METRICS constant to disable them entirely.

[kamilgrzegorczyk]

Thank you @tillkruss

Well the current situation is when the DB was pruned few days ago so that doesn't reflect the state when its actually having 500mbs.

Do you know why the number of metrics can grow so high? What are the rules for pruning/truncating the logs?

What I observed is that the issue happens mostly in the times when the admin traffic is low (like current Christmas time). The environment is dockerized and the sites work in headless mode so admin per se is not hit. Do you prune/truncate metrics on certain admin actions or cron events? (To answer the not asked question - I have server side cron running).

Maybe I will just dig it up in the source code and see what could be causing the overflow in my scenario.

[tillkruss]

Do you know why the number of metrics can grow so high? What are the rules for pruning/truncating the logs?

One HTTP request will case one metrics entry.

There is a rediscache_discard_metrics scheduled event that will delete all metrics older than WP_REDIS_METRICS_MAX_TIME (1 hour by default).

So if you WP Cron isn't working properly it may build up over time.