Re: Possible memory leak on nfsd

Re: Possible memory leak on nfsd

[Chen Chen via Bugspray Bot]

Chen Chen writes via Kernel.org Bugzilla:

Sorry for my rudeness in my previous discussion.

After switching to 6.12.4, the server stayed stable for 30 days. So whatever caused the memleak should have been resolved between 6.1.119 to 6.12.

You might want to close this bug if backport is not worthwhile.

View: https://bugzilla.kernel.org/show_bug.cgi?id=219535#c15
You can reply to this message to join the discussion.
-- 
Deet-doot-dot, I am a bot.
Kernel.org Bugzilla (bugspray 0.1-dev)

Re: Possible memory leak on nfsd

[Chuck Lever]

On 1/10/25 11:50 AM, Chen Chen via Bugspray Bot wrote:
> Chen Chen writes via Kernel.org Bugzilla:
> 
> Sorry for my rudeness in my previous discussion.
> 
> After switching to 6.12.4, the server stayed stable for 30 days.

That's good news!


> So whatever caused the memleak should have been resolved between 6.1.119 to 6.12.

That's tens of thousands of commits over two years. Unfortunately that
doesn't really tell us what the problem is.


> You might want to close this bug if backport is not worthwhile.

We need to know the exact commit that contains the fix before it can
be determined whether a backport is feasible.

Are you able to bisect between v6.1 and v6.12 ? If not, do you have
a simple, narrow reproducer that we can use to explore this ourselves?


-- 
Chuck Lever

Re: Possible memory leak on nfsd

[JJ Jordan via Bugspray Bot]

JJ Jordan added an attachment on Kernel.org Bugzilla:

Created attachment 307525
Logs and traces from Jan-18 pt1

Here are the traces from two NFS crashes that occurred this past weekend.
Both occurred in the AM (US time) on Jan 18, a few hours apart from one
another.

I followed the instructions I found on the various threads.
There was no output to `rpcdebug -m rpc -c`, not sure what I did wrong
there. The syslog ought to contain the output of sysrq-trigger, however.

The output from trace-cmd captures several days' worth of logs in either
case, but not from system boot.

The syslogs I have cut from ~one hour before the incident until it finished
shutting down prior to reboot. I have removed the output of other services.

Both are VMs on GCE running the 6.1.119 kernel from Debian bookworm (6.1.0-28)
~60Gi memory, 16 CPUs.

File: nfs-traces-250118-pt1.tar.bz2 (application/octet-stream)
Size: 4.61 MiB
Link: https://bugzilla.kernel.org/attachment.cgi?id=307525
---
Logs and traces from Jan-18 pt1

You can reply to this message to join the discussion.
-- 
Deet-doot-dot, I am a bot.
Kernel.org Bugzilla (bugspray 0.1-dev)

Re: Possible memory leak on nfsd

[JJ Jordan via Bugspray Bot]

JJ Jordan added an attachment on Kernel.org Bugzilla:

Created attachment 307526
Logs and traces from Jan-18 pt2

Part 2, see previous description

File: nfs-traces-250118-pt2.tar.bz2 (application/octet-stream)
Size: 601.99 KiB
Link: https://bugzilla.kernel.org/attachment.cgi?id=307526
---
Logs and traces from Jan-18 pt2

You can reply to this message to join the discussion.
-- 
Deet-doot-dot, I am a bot.
Kernel.org Bugzilla (bugspray 0.1-dev)

Re: Possible memory leak on nfsd

[JJ Jordan via Bugspray Bot]

JJ Jordan added an attachment on Kernel.org Bugzilla:

Comment on attachment 307525
Logs and traces from Jan-18 pt1

This was submitted in error, apologies.

File: nfs-traces-250118-pt1.tar.bz2 (application/octet-stream)
Size: 4.61 MiB
Link: https://bugzilla.kernel.org/attachment.cgi?id=307525
---
Logs and traces from Jan-18 pt1

You can reply to this message to join the discussion.
-- 
Deet-doot-dot, I am a bot.
Kernel.org Bugzilla (bugspray 0.1-dev)

Re: Possible memory leak on nfsd

[JJ Jordan via Bugspray Bot]

JJ Jordan added an attachment on Kernel.org Bugzilla:

Comment on attachment 307526
Logs and traces from Jan-18 pt2

Also submitted in error.

File: nfs-traces-250118-pt2.tar.bz2 (application/octet-stream)
Size: 601.99 KiB
Link: https://bugzilla.kernel.org/attachment.cgi?id=307526
---
Logs and traces from Jan-18 pt2

You can reply to this message to join the discussion.
-- 
Deet-doot-dot, I am a bot.
Kernel.org Bugzilla (bugspray 0.1-dev)