[SOLVED] Redis use all the CPUs

[viscid axle]

Found another solved thread here : https://discord.com/channels/564160730845151244/1220320274982109245

As he shown i also had my redis going at 90+% CPU tho i'm on 1.5.7.

Looking at the container i think the issue is that my redis container was updated by WatchTower and made the container recreate itself.
With that recreation i think appwrite send A LOT of request to redis tho all the previous data was erased because of the recreation of the redis container

A "fix" for this is to just restart all the appwrite stack (compose) and the CPU will go down.

[viscid axle]

this is 5m after the recreation

realtime is the issue i think

ID | Container Name | CPU % | MEM usage | mem% | net I/O | Storage | PID

YUP after restarting ONLY the realtime container redis is back to 1% CPU as well as the realtime container

[viscid axle]

Also since i can reproduce this state, if you want more elements i can send them (just tell me what you want and how to get it and i will do it)

[blazing gale]

If you have precise reproduction steps, it will be great and very appreciated to have a GitHub issue with the steps so the team can have a follow of it, debug and apply a fix in future releases 😄

[viscid axle]

The issue : https://github.com/appwrite/appwrite/issues/8464

[halcyon geode]

Going to need a bit more details on that, especially server specs and the exact setup you have to recreate

also remember, it's not a flat out 90% cpu usage, this is my setup rn

would highly recommend using the default containers as defined by the compose we have, if you're adding auto updates to dependent services you might run into problems. redis:7.2.4-alpine should be the latest one we officially support

1.5.8 is being deployed very soon which does include some caching fixes so you might want to keep an eye out for that

[viscid axle]

Currently this runs on a Ubuntu server 22.04 with 30GB of ram and 8 vCPU
I'm running the stack through portainer as it's easier for me to manage.

In the compose i only changed some traefik settings to make it work with mine
And the appwrite/appwrite, mariadb and docker.io/redis to latest
I also made the worker in x5 replicas but this should not be the cause

Since you told that the latest supprted was 7.2.4 i will downgrade it to see

[formal cairn]

which command shows this statistics. ?

[viscid axle]

docker stats (is also used grep to only have appwrite containers so docker stats | grep appwrite)

[formal cairn]

seems fine 1.5.10 version recreated the containers !!

[viscid axle]

redis have the high CPU usage only if the realtime container stays up while the redis container is recreated, since you updated to 1.5.10 you also updated the realtime container so not issue should appear

[blazing gale]

So is this solved now for you?

[viscid axle]

let me test i didn't try with 1.5.10

it does look like it works fine now, realtime and redis are not with a high CPU usage (it's between 1-5%) and for redis i just tested for both 7.2.4 and latest, both works (you can check on your dev env but upgrading redis is possible for 1.6 maybe)

[viscid axle]

so ye could be marked as resolved i think (i wont do it so you can close the git one at the same time)

[viscid axle]

[SOLVED] Redis use all the CPUs

[blazing gale]

Great! Just curious, is the server ARM or similar?