Samba (smbd) memory leak

Ocean · 13. Dezember 2021

After transferring a large file (>100GB) via samba, the service smbd does not release memory (uses more than 76% of memory).

Code

# ps aux --sort '%mem'
USER       PID %CPU %MEM    VSZ   RSS TTY      STAT START   TIME COMMAND
...
backup-+ 12587  0.9 76.1 6737736 4631684 ?     S    Dec12   9:38  \_ /usr/sbin/smbd --foreground --no-process-group

After restarting smbd, everything is OK

Code

# systemctl restart smbd
# ps aux --sort '%mem' | grep smbd
USER       PID %CPU %MEM    VSZ   RSS TTY      STAT START   TIME COMMAND
root      9953  0.0  0.0   6080   892 pts/0    S+   08:27   0:00 grep smbd
root      9919  0.0  0.1  46692  6184 ?        S    08:27   0:00 /usr/sbin/smbd --foreground --no-process-group
root      9918  0.0  0.1  46676  6416 ?        S    08:27   0:00 /usr/sbin/smbd --foreground --no-process-group
backup-+  9925  0.0  0.2  50792 14132 ?        S    08:27   0:00 /usr/sbin/smbd --foreground --no-process-group
root      9924  0.1  0.2  50920 14340 ?        S    08:27   0:00 /usr/sbin/smbd --foreground --no-process-group
root      9911  0.5  0.3  50572 20632 ?        Ss   08:27   0:00 /usr/sbin/smbd --foreground --no-process-group

This problem is repeated on different installations and hardware.

Version:

OMV 5.6.21-2

# samba -V

Version 4.9.5-Debian

ryecoaaron · 13. Dezember 2021

I doubt this is a memory leak. It is probably just Linux file cache - https://www.linuxatemyram.com/. Did you look at the output of free -m? If it truly is a memory leak, OMV is using Debian provided packages since OMV is debian. So, a bug report would need to be filed with Debian to get a fix.

Ocean · 18. Dezember 2021

This is not a cache

Code

# free -m
              total        used        free      shared  buff/cache   available
Mem:           3933        3062         284           4         586         637
Swap:          4093        2197        1896

used 3062...

and begins to use swap...

ryecoaaron · 18. Dezember 2021

Zitat von Ocean

This is not a cache

What else is running on the box? Samba would be causing lots of people problems if there was a memory leak.

Ocean · 18. Dezember 2021

Zitat von ryecoaaron

What else is running on the box? Samba would be causing lots of people problems if there was a memory leak.

Zabbix Agent 5.0. I will try to disable the agent.

Ocean · 18. Dezember 2021

Zitat von ryecoaaron

What else is running on the box? Samba would be causing lots of people problems if there was a memory leak.

I was mistaken in the previous post. The zabbix-agent is running on only one server, but the problem is present on two servers.

The problem appears only when a large file is transferred (about 100 GB) and at a speed> 70 mb / s

I will watch the problem.

rucknapucknavitz · 26. Juni 2022

Wonder if you’re still having this issue? It started for me under OMV6 with a oom-killer. So I increased ram from 2Gbto 4Gb. Still getting the error, with smbd taking 3.5Gb after every restart.

fvlaicu · 5. November 2022

It started for me for a while now, it happens when I transfer large files i.e.:

It happens when I try to backup a disk using clonezilla, now it happened after I transfered a 20GB file.

Look at the screenshot below: until last night around 10 PM memory usage was kinda flat, at 10 I transfered the 20GB file to the server.

It went bonkers and got oom killed around 4 AM, after samba was killed memory usage goes flat again to levels similar to before 10 PM.

The fact that it was killed at 4 AM is also telling that it happened around the time cron.daily runs. Perhaps, cron ran and pushed the memory usage up, but the oomkiller chose samba.

Still, the fact that samba didn't free the memory after 6 hours is weird :).

I can say that samba was killed because I have the log here too below the screenshot.

Code

systemctl status smbd
● smbd.service - Samba SMB Daemon
     Loaded: loaded (/lib/systemd/system/smbd.service; enabled; vendor preset: enabled)
     Active: failed (Result: oom-kill) since Sat 2022-11-05 04:02:52 EET; 4h 6min ago
       Docs: man:smbd(8)
             man:samba(7)
             man:smb.conf(5)
    Process: 2656 ExecStart=/usr/sbin/smbd --foreground --no-process-group $SMBDOPTIONS (code=killed, signal=TERM)
   Main PI[tt]<span id="selection-marker-1" class="redactor-selection-marker woltlab-bbcode-marker"></span>[/tt]D: 2656 (code=killed, signal=TERM)
     Status: "smbd: ready to serve connections..."
        CPU: 1min 30.977sNov 05 04:02:52 omv systemd[1]: smbd.service: A process of this unit has been killed by the OOM killer.
Nov 05 04:02:52 omv systemd[1]: smbd.service: Failed with result 'oom-kill'.
Nov 05 04:02:52 omv systemd[1]: smbd.service: Consumed 1min 30.977s CPU time.

Alles anzeigen

upsangel · 3. Oktober 2023

Zitat von fvlaicu

It happens when I try to backup a disk using clonezilla,

I happens to me exactly the same when I doing a smb backup to OMV with clonezilla. Seems it's a repeated problem.

Samba (smbd) memory leak

Ocean 13. Dezember 2021

Jetzt mitmachen!

Tags