Re: Filesystem lockup during backup

[Qu Wenruo <quwenruo.btrfs@gmx.com>]

在 2025/10/20 08:25, Torbjörn Jansson 写道:
> On 2025-10-19 22:42, Qu Wenruo wrote:
>>
>>
>> 在 2025/10/19 20:43, Torbjörn Jansson 写道:
>>> Hello.
>>>
>>> i have a btrfs filesystem on two 18tb disks that i use as backup 
>>> destination for my proxmox cluster.
>>> the filesystem is using btrfs raid1 mirroring and is exported over 
>>> nfs to the other nodes.
>>>
>>> because this is used primarily for backups there are periods of heavy 
>>> writes (several backups running at the same time) and when this 
>>> happens it is very likely the filesystem and nfsd locks up completely.
>>> this then starts a chain reaction due to the default hard mount 
>>> blocking processes then eventually ceph also becomes unhappy and then 
>>> the vms goes down.
>>>
>>> below is the hung task output from dmesg on the computer with the disks.
>>>
>>> any idea whats going on and what i can do about it?
>>>
>>>
>>>
>>> [1560204.654347] INFO: task nfsd:5136 blocked for more than 122 seconds.
>>> [1560204.654351]       Tainted: P           O       6.14.11-2-pve #1
>>
>> v6.14 is EOL, you're completely on the vendor to provide any fix/ 
>> backport.
>>
>> Recommended to go either LTS kernels or latest upstream one.
>>
> 
> ok, not sure i can easily fix that.
> i can go back to 6.8 but i think thats not an lts release.
> i believe proxmox just released a 6.17 kernel but thats still in the 
> test repo and if i enable that i might accidentally end up with test 
> versions of other packages and i dont want that.

In that case, I'm afraid you have to ask PVE for support.

Neither 6.8 bir 6.14 are supported upstream, thus there must be some PVE 
specific backports there.

Or you may want to explore how to install upstream kernels on your distro.
> 
> 
>>> [1560204.654353] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" 
>>> disables this message.
>>> [1560204.654355] task:nfsd            state:D stack:0     pid:5136 
>>> tgid:5136  ppid:2      task_flags:0x200040 flags:0x00004000
>>
>> The only message? No more other tasks?
>>
> 
> the same (or very similar) message is repeated 10 times, all related to 
> nfsd and then it stops with:
> "Future hung task reports are suppressed, see sysctl 
> kernel.hung_task_warnings"
> 
> if you want i can provide all 10

Please provide full dmesg when reporting kernel related bugs.

It may seem the same to you, but a lot of time important details are 
hidden inside the full dmesg.

Thanks,
Qu

> 
> 
>>> [1560204.654361] Call Trace:
>>> [1560204.654363]  <TASK>
>>> [1560204.654366]  __schedule+0x466/0x1400
>>> [1560204.654376]  schedule+0x29/0x130
>>> [1560204.654381]  io_schedule+0x4c/0x80
>>> [1560204.654387]  folio_wait_bit_common+0x122/0x2e0
>>> [1560204.654393]  ? __pfx_wake_page_function+0x10/0x10
>>> [1560204.654400]  __folio_lock+0x17/0x30
>>
>> This hangs at folio locking, which normally means another process has 
>> locked the data folio, but no more hanging messages to further debug it.
>>
>> Thanks,
>> Qu
>>
>>> [1560204.654404]  extent_write_cache_pages+0x36e/0x7f0 [btrfs]
>>> [1560204.654559]  btrfs_writepages+0x75/0x130 [btrfs]
>>> [1560204.654703]  do_writepages+0xde/0x280
>>> [1560204.654710]  ? __pfx_ip_finish_output+0x10/0x10
>>> [1560204.654715]  ? wbc_attach_and_unlock_inode+0xd1/0x130
>>> [1560204.654721]  filemap_fdatawrite_wbc+0x58/0x80
>>> [1560204.654726]  ? __ip_queue_xmit+0x19b/0x4e0
>>> [1560204.654731]  __filemap_fdatawrite_range+0x6d/0xa0
>>> [1560204.654744]  filemap_fdatawrite_range+0x13/0x30
>>> [1560204.654748]  btrfs_fdatawrite_range+0x28/0x70 [btrfs]
>>> [1560204.654889]  start_ordered_ops.constprop.0+0x4e/0x90 [btrfs]
>>> [1560204.655029]  btrfs_sync_file+0xa9/0x610 [btrfs]
>>> [1560204.655159]  ? list_lru_del_obj+0xad/0xe0
>>> [1560204.655168]  vfs_fsync_range+0x42/0xa0
>>> [1560204.655174]  nfsd_commit+0x9f/0x180 [nfsd]
>>> [1560204.655275]  nfsd4_commit+0x60/0xa0 [nfsd]
>>> [1560204.655367]  nfsd4_proc_compound+0x3ad/0x760 [nfsd]
>>> [1560204.655427]  nfsd_dispatch+0xce/0x220 [nfsd]
>>> [1560204.655486]  svc_process_common+0x464/0x6f0 [sunrpc]
>>> [1560204.655553]  ? __pfx_nfsd_dispatch+0x10/0x10 [nfsd]
>>> [1560204.655611]  svc_process+0x136/0x1f0 [sunrpc]
>>> [1560204.655675]  svc_recv+0x7bb/0x9a0 [sunrpc]
>>> [1560204.655741]  ? __pfx_nfsd+0x10/0x10 [nfsd]
>>> [1560204.655798]  nfsd+0x90/0xf0 [nfsd]
>>> [1560204.655852]  kthread+0xf9/0x230
>>> [1560204.655855]  ? __pfx_kthread+0x10/0x10
>>> [1560204.655858]  ret_from_fork+0x44/0x70
>>> [1560204.655862]  ? __pfx_kthread+0x10/0x10
>>> [1560204.655864]  ret_from_fork_asm+0x1a/0x30
>>> [1560204.655871]  </TASK>
>>>
> 
> 
>