Possible memory leak in mate-panel?

Hi

My desktop machine, with 8GB of RAM was running sluggish.
Ubuntu MATE 16.04.1 LTS w/ MATE 1.12.1
I've noticed mate-panel was eating a ton of memory:

I've rebooted and here's what it was using right after startup:

Anyone else noticed this? The system had only about 10 days of uptime.

[email protected]:~/Desktop$ last -n 100 | grep reboot | head -n 2
reboot   system boot  4.4.0-31-generic Sat Jul 23 23:15   still running
reboot   system boot  4.4.0-28-generic Wed Jul 13 19:25 - 23:15 (10+03:50)

Anyone know a less violent way to recover the resources that won't screw up my customized panel layout?

Hi @ouroumov,

I haven’t noticed anything unusual myself, have you considered upgrading to Mate 1-4?:

Could be. I see mate-panel using up more and more memory day by day. It is not something you would notice unless you run for several days without rebooting.

1 Like

I updated and rebooted this morning, so here’s the current state (about 32M).

Code:
top -p 4218

  PID USER      PR  NI    VIRT    RES    SHR S  %CPU %MEM     TIME+ COMMAND     
 4218 m1        20   0  640796  32432  24412 S   0.0  0.1   0:03.08 mate-panel

I guess you show about the same after reboot?

I’ll take another look tomorrow and see if it changed.

1 Like

After about 28 hours: (AMD64 system)

  PID USER      PR  NI    VIRT    RES    SHR S  %CPU %MEM     TIME+ COMMAND     
 3397 superman  20   0 1043632  75280  25980 S   0.0  0.5   0:19.01 mate-panel  

My servers do not show any memory increase so it might be related to how much you use the mate-panel.

1 Like

Looks like a good cause. It may be a particular applet and I’ve added a few.

top - 06:21:57 up 0 min,  2 users,  load average: 0.53, 0.20, 0.07
  PID USER      PR  NI    VIRT    RES    SHR S  %CPU %MEM     TIME+ COMMAND
 3737 bill      20   0  635908  32208  22644 S   0.0  0.1   0:00.22 mate-panel
1 Like

I’ve considered it, then I decided against it, mainly for support reasons.
But that’s not important either way, MATE 1.12.1 is the one shipping with the LTS so if there happens to be a problem, it should be fixed rather than worked around.

My work machine, with 5 days of uptime:

  PID USER      PR  NI    VIRT    RES    SHR S  %CPU %MEM     TIME+ COMMAND     
 1371 ouroumov  20   0  867572  42948  26452 S   0,0  0,3   0:10.72 mate-panel  

I was wondering about that, is it possible a specific applet is to blame?
I’m using the same set of applets on all my machines:

  • Default everything (standard layout)
  • Added disk mounter applet to bottom panel
  • Added system monitor applet to top panel, and set it to watch CPU, RAM, NET and DISK I/O
1 Like

I’ve added weather with custom animated GIF radar and a command output every 10 seconds for cat /var/run/reboot-required (I know, seems a waste). Claws mail is active in the notification area. The rest are all static launchers.

Your 5-day work machine indicates this may not be a linear creep but a burst happening all at once. Your first values were rather alarming.

1 Like

The Drawer applet appears to be one culprit. Running something from the Drawer adds memory to mate-panel. There may be more because something else is making the memory values jump up a lot more.

Well, I just went back to square one. Popped the circuit breaker and took out my box.

I did look at it last night and did see a small (insufficient) rise in ram.

Starting over :frowning:

Edit:
Just for reference, here's my panel layout (no dock)

I also had to restart from a power failure but up to 1d 18h and RES is up 40%.

Two days now and nothing to report.

I’m shy of 4 days and it’s definitely heading upwards but not at anything alarming. It bounced up and down a tiny bit along the way.

32.2M START (RES = 32208)
53.6M 3 days 18:21 (RES = 53584)

Day 6 after the reboot and I think this is happening again:

Note the CPU usage of the parent mate-panel process.
Also note the SWAP usage: this system has an SSD with vm.swappiness=1 added to sysctl.conf.

I now think this might occur because of a weird interaction with the Eye Of Mate image viewer.
I have a massive slideshow running on one of my screens, and the CPU spikes are following the image transitions.

I will attempt to reproduce on my work machine on Monday and if I manage to do so, I will create steps to reproduce for a future bug report.

Edit: in the mean time I'm gathering some logs of top running in batch mode. I will let it run for a few hours, monitoring the eom and mate-panel memory usage and pause eom for a while to see if the memory increase correlates with the period of activity of eom.

1 Like

OK so, pardon my poor chart-making skills...
Left scale is memory, right scale is CPU.

I'd like to post the full log but it's too big for pastebin et al...
Measurements taken at 10 seconds interval for about 13 hours.
@Wimpy, @lah7, what do you think?

Edit: should have killed eom before I left for the week end.
New record:

2 Likes

Filed two separate bug reports:

against mate-panel : https://bugs.launchpad.net/ubuntu-mate/+bug/1608599 (with steps to reproduce)
against eom : https://bugs.launchpad.net/ubuntu-mate/+bug/1608603

4 Likes

So guys, was anyone able to reproduce?

The last I had logged was 5d 6h 26m and mate-panel RES was up to 182% of original (58708/32208).

Between VirtualBox getting super-sluggish and Caja crashing on ssh/sftp (I may post about these) I had to abandon the effort and reboot. VirtualBox recovers but the Caja crashes remain frequent. :disappointed:

When I first boot up my Lenovo Thinkpad T400, memory usage for Ubuntu MATE 16.04 32-bit at idle is anywhere from 700-900 MB. This is incredible! Ubuntu MATE 14.04 32-bit idles at 260 MB. I have even unchecked many startup programs and it hasn’t helped. Does anyone know what is causing this?

I don’t know why your memory usage is so high.

I have UM 14.04 and UM 16.04 on the same machine. I get the following results (same start-up applications):

UM 14.04 64-bit uses 344 MiB
UM 16.04 64-bit uses 542 MiB