Re: OSD blocked for more than 120 seconds

[Martin Mailand <martin@tuxadero.com>]

Hi Sage,
the hang was on a btrfs, I do not have a fix for that.

The 'filestore flusher = false' does fix the ext4 problems, which where 
reported from Christian, but this option has quite an impact of the osd 
performance.
The '-o noauto_da_alloc' option did not solve the fsck problem.

Best Regards,
  Martin


Sage Weil schrieb:
> On Mon, 17 Oct 2011, Martin Mailand wrote:
>> Am 17.10.2011 11:40, schrieb Christian Brunner:
>>> Our bugreport with RedHat didn't make any progress for a long time,
>>> but last week RedHat made two sugestions:
>>>
>>> - If you configure ceph with 'filestore flusher = false', do you see
>>> any different behavior?
>>> - If you mount with -o noauto_da_alloc does it change anything?
>> Hi,
>> after a quick test I think 'filestore flusher = false' did the trick.
>> What does it do?
> 
> It fixes your hang (previous email), or the subsequent fsck errors?
> 
> When filestore flusher = true (default), after every write the fd is 
> handed off to another thread that uses sync_file_range() to push the data 
> out to disk quickly before closing the file.  The purpose is to limit the 
> latency for the eventual snapshot or sync.  Eric suspected the handoff 
> between threads may be what was triggering the bug in ext4.
> 
> sage