-
Notifications
You must be signed in to change notification settings - Fork 582
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
API memory leak with signalilo #8290
Comments
Hello @yoshi314 and thank you for reporting! Please provide instructions how to reproduce this w/o K8s, LXC or signalilo. At best just Icinga 2, a shell and curl. Best, |
well, i have no clue how to install signalilo outside of k8s, but you might simply deploy it in minikube with helm. use provided vales.yaml and do ( in git checkout directory )
i am not sure how to use helm with minikube, but i'll take a look. |
technically signalilo only sends passive check every minute or so. so any other API client will likely do the trick. |
Hello again @yoshi314! Please could you try #7864? https://git.icinga.com/packaging/deb-icinga2/-/jobs/66019 / Job artifacts / Download Best, |
Note: Caught the symptom that parallel |
ok, i'll try this in a moment and report in a few days. (leak is pretty slow) |
i am testing now with this version. i'll keep you posted. |
it's not looking promising so far, but i'll wait and see if it oom's. That "W" in the plot is the moment of upgrade. edit: it looks more or less stable for now. but i'll check back in few hours. |
Unfortunately i am still hitting the memory leak. I'll disable signalilo and keep this buiild running. Test build was installed at 8:10 and it had two reloads in the meantime. The dip at ~17:00 is when icinga2 got oom'd. |
This is the situation after i disabled signalilo (~20:00). As you an see, things have evened out almost instantly. |
No idea if this is related, it's just a post by a random user (me) getting the impression, that the central hub icinga2 process looks like it's blowing up like Akira. Maybe to quote from that post: snip 8<
We are monitoring about 40 systems and have about 450 checks going. Last week the system started whining abour swap space running out. So I doubled the swap space. Now swap space is used up again. And the memory used by the main icinga only keeps on growing. This is the central monitoring hub of ours where all satellites are connecting to and reporting their stati. The massive growth of the memory consumed by the main icinga2 process only seems to have started since a week or so and does not correlate to anything I’m aware of (except restarting the host server around the same time, but that correlation doesn’t make sense to me). So, is there maybe a memory leak at work in the main icinga2 process? Is that much memory consumption normal? Any hints or clues? snap 8< |
I would check if you have some kind of passive check or api client that you enabled recently. |
The signalilo use case is basically to poke icinga2 once per minute with a heartbeat check. It may occasionally add new checks from prometheus' alertmanager but i have not setup that yet. It's worth noting that i am not getting such issues with Dashing for icinga2 (i used to, maybe two years ago). Perhaps the client itself is misbehaving somehow? |
I've tried to install those, to see if it maybe fixes my problem, however, my system is an Ubuntu bionic system, and so I am getting:
You don't have by chance packages for bionic lying around somewhere? The icinga on Ubuntu bionic has these boost dependencies:
|
You need these ones: https://git.icinga.com/packaging/deb-icinga2/-/jobs/66027 |
So Icinga is managed by systemd, right? |
yes, it is. i might be able to switch to classic init, since it's debian. |
Thanks a lot @Al2Klimov - unfortunately that doesn't fix the leak/huge memory consumption. |
release/package 2.12.2-1.bionic seems to have fixed my problem. To all of you who are improving/fixing icinga: thanks a lot! |
i have a suspicion that 2.12.2 might have fixed it, since my cluster survived so far. I'll be testing it a bit more over the week. |
I'm afraid, that was only temporary. Icinga2 is still using huge amounts of memory:
|
ref/NC/700554 |
A fix for this (on the side of signalilo) is available which is already merged but not released: vshn/signalilo#63 |
@dgoetz are there any *deb packages or executables for me to test? |
Our customer is using docker images. Unfortunately not sure about other installation methods as I have not worked with signalilo personally. |
OK folks, release/package/version 2.12.4-1.bionic seems to have fixed the memory leak problem for me: icinga is not using memory like crazy any more. Thanks a lot to all involved!!!* @dgoetz @Al2Klimov @N-o-X ! @yoshi314 maybe you want to recheck whether your problem ist still persisting, otherwise this ticket could be closed? |
i haven't seen the leak in a long while, but i will keep an eye out |
considering fixed for now. haven't seen a memleark in a while. |
Describe the bug
I use icinga2 on lxc, and signalilo runs in k8s.
I have installed signalilo as follows :
checkout https://github.com/appuio/charts/
cd signalilo
edit values.yaml to your liking, e.g
Define a passive check in icinga2
To Reproduce
Disabling signalilo makes icinga2 work with no issues.
Expected behavior
Icinga2 works.
Screenshots
https://imgur.com/a/zKZlRUm
You can notice the OOM in the flat area of the plot.
Your Environment
icinga2 --version
):icinga2 feature list
):The text was updated successfully, but these errors were encountered: