Hi Dan,

when the warning popped up the system was running for months without any
issues. The cluster is running 18.2.7 and uses 2 active MDS.
I started monitoring the object and noticed the amount of keys going beyond
the 200k threshold multiple times.
Unfortunately we had an outage Monday night (not caused by Ceph :) ), so we
had to restart every service.
Since then I didn't observe any openfiles objects with more than 200k keys.
I set up a monitoring for all files though, so we should be alerted when
the issue arises again.

Thank you!
Philipp

On Wed, Nov 19, 2025 at 8:43 AM Dan van der Ster <[email protected]>
wrote:

> Hi Philipp,
>
> This looks like https://tracker.ceph.com/issues/61950, which was
> closed as not reproducible.
>
> Do you have any more background context about that cluster that might
> give a clue how this happened? maybe a recent daemon
> upgrade/restart/...?
>
> Cheers, Dan
>
> On Mon, Nov 17, 2025 at 8:28 AM Philipp Hocke <[email protected]>
> wrote:
> >
> > Hi,
> >
> > last week we had a "large omap" warning in one of our clusters. The
> > affected object was
> >
> > 025-11-09T23:31:24.799+0000 7f70131b3640  0 log_channel(cluster) log
> > [WRN] : Large omap object found. Object:
> > 2:ddb71591:::mds1_openfiles.d:head PG: 2.89a8edbb (2.b) Key count:
> > 200001 Size (bytes): 12122520
> >
> > I wanted to get a deeper understanding on what triggers this specific
> issue
> > and had a look at the code - I'm not that deep into C++, so please,
> correct
> > me if I'm wrong.
> >
> > If I understand correctly, LARGE_OMAP warnings are only generated (and
> > cleared) during deep-scrubs. If an object is found with more keys
> > than osd_deep_scrub_large_omap_object_key_threshold, it will trigger the
> > warning. In this case
> >
> > mds1_openfiles.d had a Key count of 200001 while this specific option
> > got deep-scrubbed and that triggered the warning
> >
> > The MDS should create a new openfiles fragment when the threshold is
> reached.
> >
> > My question is: How exactly does the MDS end up with an openfiles
> > segment with more than 200000? Shouldn't the MDS create a new segment
> > as soon as the previous one reaches the configured limit?
> >
> > If this is considered a bug, I'm happy to open a report.
> >
> >
> > Best regards
> >
> > Philipp
> >
> > --
> >
> > Philipp Hocke
> > Leiter Linux Systemadministration / Head Of TechOps
> > CM4ALL GmbH
> > Im Mediapark 6A - 50670 Köln / Cologne
> > Phone +49-(0)221-6601-0 <https://outlook.office.com/mail/sentitems>
> > Fax +49-(0)221-6601-1011 <https://outlook.office.com/mail/sentitems>
> > E-Mail: [email protected]
> > Internet: www.cm4all.com
> > _______________________________________________
> > ceph-users mailing list -- [email protected]
> > To unsubscribe send an email to [email protected]
>
>
>
> --
> Dan van der Ster
> Ceph Executive Council | CTO @ CLYSO
> Try our Ceph Analyzer -- https://analyzer.clyso.com/
> https://clyso.com | [email protected]
>


-- 

Philipp Hocke
Leiter Linux Systemadministration / Head Of TechOps
CM4ALL GmbH
Im Mediapark 6A - 50670 Köln / Cologne
Phone +49-(0)221-6601-0 <https://outlook.office.com/mail/sentitems>
Fax +49-(0)221-6601-1011 <https://outlook.office.com/mail/sentitems>
E-Mail: [email protected]
Internet: www.cm4all.com

Philipp Hocke
Leiter Linux Systemadministration / Head Of TechOps
CM4ALL GmbH
Im Mediapark 6A - 50670 Köln / Cologne
Phone +49-(0)221-6601-0 <https://outlook.office.com/mail/sentitems>
Fax +49-(0)221-6601-1011 <https://outlook.office.com/mail/sentitems>
E-Mail: [email protected]
Internet: www.cm4all.com
_______________________________________________
ceph-users mailing list -- [email protected]
To unsubscribe send an email to [email protected]

Reply via email to