shlinkio
High memory usage? Over 200MB for mostly-idle container #59
How Shlink is set-upContainerized, version 3.1.2, hosted on ubuntu docker host. Database is mysql on another server. SummaryCurious as to why there is high-RAM usage. I have 2 instances and both use a little above/below 200MB on average. Wondering why this is. Screenshot below is just shy of 200MB, but I've seen it go past that mark. Thanks!
|
Replies: 3 comments 6 replies
|
It's hard to tell, but I imagine openswoole (the runtime used to serve Shlink on the docker image) reserves some memory by default. I will try to investigate a bit further how this works, and if it's configurable, but if your instances don't have a lot of traffick, maybe you can try reducing the amount of task and web workers, as described here It uses 16 of each by default. I would try with 8 or so and see if it helps. |
|
Awesome. I'll check this out and report back. Thanks!
------ Original Message ------
From "Alejandro Celaya" ***@***.***>
To "shlinkio/shlink" ***@***.***>
Cc "Ryan Peters" ***@***.***>; "Author"
***@***.***>
Date 8/11/2022 2:55:31 PM
Subject Re: [shlinkio/shlink] High memory usage? Over 200MB for
mostly-idle container (Discussion shlinkio/shlink#1505)
It's hard to tell, but I imagine openswoole <https://openswoole.com/>
(the runtime used to serve Shlink on the docker image) reserves some
memory by default.
I will try to investigate a bit further how this works, and if it's
configurable, but if your instances don't have a lot of traffick, maybe
you can try reducing the amount of task and web workers, as described
here <https://shlink.io/documentation/environment-variables/#server>
It uses 16 of each by default. I would try with 8 or so and see if it
helps.
—
Reply to this email directly, view it on GitHub
<https://github.com/shlinkio/shlink/discussions/1505#discussioncomment-3378284>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AANN2X7DLXX2H3UIPEPT5BDVYVECHANCNFSM56JGWNHQ>.
You are receiving this because you authored the thread.Message ID:
***@***.***>
… |
|
Will be nice to see something in this area, the memory usage for the api in really high when is IDLE for almost 99% of time: 
|
|
This is a very old thread. Shlink is no longer using openswoole, but RoadRunner. RoadRunner will reserve around 50/60 MB of memory per worker, so the best approach to reduce that is reduce the amount of workers if you are not having a lot of traffic. In any case, 250MB doesn't seem terrible to me 😅 |
|
I was investigating what looked like a memory leak some time ago. You can see my notes there shlinkio/shlink#2433 (comment) |
|
Yeap, I see here https://github.com/roadrunner-server/roadrunner/blob/3c6ee7d781f5d3f3f9d0678fc4c1a1719d6167f8/.rr.yaml#L340 will use a bit of memory for worker. TBH 160MB is high for IDLE, look how amazing the admin are using only 5MB. |
|
After change this: WEB_WORKER_NUM=1
TASK_WORKER_NUM=1Worked, I think this can be close! |
That's only the maximum memory you allow for a worker before it gets killed, not the minimum. Shlink would require an in-depth memory profiling to find out where is the memory going. Would be great if you want to give that a chance.
It makes no sense to compare those two really. |
|
Good point about the memory, I will try use this to force some SWAP as part of this is a possible dirt pages: deploy:
resources:
limits:
cpus: '2.00'
memory: 256MLast time I worked with PHP was 5.3-7-ish |
After change this:
Worked, I think this can be close!