Services slowly eating up RAM

[chri5]

My services are slowly eating up more and more ram. After a reboot I usually have over 16gb ram free (32 total). But as the days go by my services slowly eat up ram despite me not adding to or changing the existing services. Is this usual?

I should add I'm using, plex, nextcloud & traefik. 1 VM running ubuntu server with nginx, and a samba share.

 

[LarsR]

Have you installed any apps or truecharts as an additional repository?
Right now the catalogue sync data is stored in ram and counts to system memory. It's syncing the previous and the current catalogue data.
That could be 1-2gb of data for each sync when truecharts is added.
If you don't have any apps installed it could be a memory leak.

 

[chri5]

Have you installed any apps or truecharts as an additional repository?
Right now the catalogue sync data is stored in ram and counts to system memory. It's syncing the previous and the current catalogue data.
That could be 1-2gb of data for each sync when truecharts is added.
If you don't have any apps installed it could be a memory leak.

Just Traefik and nextcloud from the truecharts core repo. Plex is from the official repo.

 

[sretalla]

a samba share

SMB counts in services and it does do caching of its own, so can grow in memory use over time.

 

[bomg]

@chri5, you might be interested in this thread discussing memory bloat issues related to the app catalog (pages 1-2) and to keeping the web dashboard open (pages 2-3).

 

[anodos]

SMB counts in services and it does do caching of its own, so can grow in memory use over time.

There is limit to how large smbd process should grow. Memory usage will be visible in `top` and `ps` output and associated process will be `smbd`.

 

[MountainMan]

This is a frustrating issue. I *feel* like there there are multiple issues at play here with middlewared -- which has a variety of responsibilities it seems. I see it's memory footprint endlessly grow, and CPU use bumping up over time after some yet-unknown event(s).

In looking with pmap, the anon memory blocks endlessly increase and become the vast majority of the memory by middlewared for me over time. I made a script to track that over time which I've just started up. Plan on rebooting, not using the UI at all for a day, then leave a browser on the dashboard for a day, normal use for a day, etc. Horrible way to try and identify things, but the tickets in JIRA seem pretty stalled.

Just seems odd more people don't seem to experience it, as I have a pretty simple configuration with just a few pools, SMB, and one official app (Plex) running.

 

[Trevor68]

I too have a quite simple setup and only 6 apps running. When I started with SCALE at the RC point I could go a month easy without using up all the RAM. Now it's about a week.

 

[Trevor68]

Ok, it looks like I lose a chunk of RAM at 10pm every night? How would I track that down?

 

[asc72]

I am seeing the same issue. Services continue to grow until I restart the server. I have 5 apps installed, no VMs.

 

[bacardi8]

Same here.
Plex and Emby is the only apps that I have installed.
Before I never see that, so maybe since the latest update this is happening..??

Until now I just restart if I want to bring it down to normal again, but for now I will try to let it run,
just to see how high it goes.

A question here is, will it go higher than 50%, and will it automatically drop/free up Ram, if these are needed for other tasks...???
this would then actually be normal behavior as far as I know.

 

[Botched1]

Same here. I know that doesn't help, in and of itself.

On reboot services use ~5GB on my system.
After 25 days uptime they were up to 23GB.

Just reboot, now back to 5GB.

I wouldn't be surprised by SOME increase over time as I am running 1 VM (2GB ram allocated) and 1 container. But 5GB to 20+GB seems like a lot.

Memory Graph - I don't see one that specifically shows "services":

 

[Trevor68]

Mine grows like clockwork at 9pm everyday, guess I will need to track down what it's doing at that time somehow.

 

[bacardi8]

Mine after 2 days

 

[Trevor68]

I see a "User" CPU spike and corresponding drop in free RAM at 9PM each night, I have no system scheduled tasks, so either the Truecharts Catalog is updating then, or one of my apps are.

 

[pabloalcantara]

Probably is some kind of middleared memory leak.
I have created a "solution", it´s not the best o elegant, but it´s working.
I have created a simple script that runs "systemctl restart middlewared" every 2 days.

 

[chri5]

After each reboot you can see free memory decline and used increase.