[Gluster-users] [Gluster-devel] gluster 10.3: task glfs_fusenoti blocked for more than 120 seconds

[Hu Bert]

Hi Mohit,
strange, after the latest kernel update (5.10.0-22-amd64 #1 Debian
5.10.178-3) and with gluster 10.4 running, i got this message again
today. Result: for a certain time period e.g. 'df -h' hangs. And as
gluster is the only thing (at least that i'm aware of) that uses FUSE
i thought there might... ;-)

Maybe i'll have to check if there's anything that can be done on
kernel/system side. Searching... :-)


Thx,
Hubert


Am Di., 2. Mai 2023 um 14:28 Uhr schrieb Mohit Agrawal <moagrawa at redhat.com>:
>
> I don't think the issue is on gluster side, it seems the issue is on kernel side (possible deadlock in fuse_reverse_inval_entry)
> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=bda9a71980e083699a0360963c0135657b73f47a
>
> On Tue, May 2, 2023 at 5:48 PM Hu Bert <revirii at googlemail.com> wrote:
>>
>> Good morning,
>>
>> we've recently had some strange message in /var/log/syslog.
>>
>> System:
>> debian bullseye, kernel 5.10.0-21-amd64 and 5.10.0-22-amd64
>> gluster 10.3
>>
>> The message look like:
>>
>> Apr 27 13:30:18 piggy kernel: [24287.715229] INFO: task
>> glfs_fusenoti:2787 blocked for more than 120 seconds.
>> Apr 27 13:30:18 piggy kernel: [24287.715327]       Not tainted
>> 5.10.0-22-amd64 #1 Debian 5.10.178-3
>> Apr 27 13:30:18 piggy kernel: [24287.715419] "echo 0 >
>> /proc/sys/kernel/hung_task_timeout_secs" disables this message.
>> Apr 27 13:30:18 piggy kernel: [24287.715575] task:glfs_fusenoti
>> state:D stack:    0 pid: 2787 ppid:     1 flags:0x00000000
>> Apr 27 13:30:18 piggy kernel: [24287.715734] Call Trace:
>> Apr 27 13:30:18 piggy kernel: [24287.715847]  __schedule+0x282/0x870
>> Apr 27 13:30:18 piggy kernel: [24287.715959]  schedule+0x46/0xb0
>> Apr 27 13:30:18 piggy kernel: [24287.716073]
>> rwsem_down_write_slowpath+0x257/0x4d0
>> Apr 27 13:30:18 piggy kernel: [24287.716194]
>> fuse_reverse_inval_entry+0x3b/0x1e0 [fuse]
>>
>> etc. Full excerpt here: https://pastebin.com/6gDHgh16
>>
>> This may cause our application to "hang". In the latest release i read
>> something about fuse:
>> https://docs.gluster.org/en/latest/release-notes/10.4/
>> "Fix fuse concurrency problems"
>>
>> but i checked the tickets and wasn't able to find something.
>>
>> However, i upgraded gluster client 10.3 -> 10.4, hoping that this
>> fixes the hang-issue.
>>
>> Has anyone seen these messages before?
>>
>>
>> Best regards,
>> Hubert
>> -------
>>
>> Community Meeting Calendar:
>> Schedule -
>> Every 2nd and 4th Tuesday at 14:30 IST / 09:00 UTC
>> Bridge: https://meet.google.com/cpu-eiue-hvk
>>
>> Gluster-devel mailing list
>> Gluster-devel at gluster.org
>> https://lists.gluster.org/mailman/listinfo/gluster-devel
>>