Hi guys,
My Ceph release is Quincy 17.2.5. I need to change the master zone to
decommission the old one and upgrade all zones. I have separated the client
traffic and sync traffic in RGWs, meaning there are separate RGW daemons
handling the sync process.
I encountered an issue when trying to sync one of the zones in the
zonegroup. The data sync is proceeding fine, but I have an issue with the
metadata sync. It gets stuck behind on a shard. Here is the output
from radosgw-admin
sync status:
metadata sync syncing
full sync: 1/64 shards
full sync: 135 entries to sync
incremental sync: 63/64 shards
metadata is behind on 1 shard
behind shards: [0]
In the RGW log, I see this error:
2024-12-15T21:30:59.641+0000 7f6dff472700 1 beast: 0x7f6d2f1cf730:
172.19.66.112 - s3-cdn-user [15/Dec/2024:21:30:59.641 +0000] "GET
/admin/log/?type=data&id=56&marker=00000000000000000000%3A00000000000000204086&extra-info=true&rgwx-zonegroup=7c01d60f-88c6-4192-baf7-d725260bf05d
HTTP/1.1" 200 44 - - - latency=0.000000000s
2024-12-15T21:30:59.701+0000 7f6e44d1e700 0 meta sync: ERROR: full_sync():
RGWRadosGetOmapKeysCR() returned ret=-2
2024-12-15T21:30:59.701+0000 7f6e44d1e700 0 RGW-SYNC:meta:shard[0]: ERROR:
failed to list omap keys, status=-2
2024-12-15T21:30:59.701+0000 7f6e44d1e700 0 meta sync: ERROR:
RGWBackoffControlCR called coroutine returned -2
2024-12-15T21:31:00.705+0000 7f6e44d1e700 0 meta sync: ERROR: full_sync():
RGWRadosGetOmapKeysCR() returned ret=-2
2024-12-15T21:31:00.705+0000 7f6e44d1e700 0 RGW-SYNC:meta:shard[0]: ERROR:
failed to list omap keys, status=-2
2024-12-15T21:31:00.705+0000 7f6e44d1e700 0 meta sync: ERROR:
RGWBackoffControlCR called coroutine returned -2
I’ve tried the following steps:
- Changed the PG number of the metadata pool to force a rebalance, but
everything was fine.
- Ran metadata sync init and tried to run it again.
- Restarted RGW services in both the zone and the master zone.
- Created a user in the master zone to ensure metadata sync works, which
was successful.
- Checked OSD logs but didn’t see any specific errors.
- Attempted to list metadata in the pool using rados ls -p
s3-cdn-dc07.rgw.meta, but got an empty result.
- Compared the code for listing OMAP keys between Quincy and Squid
versions; there were no specific changes.
I’m looking for any advice or suggestions to resolve this issue.
_______________________________________________
ceph-users mailing list -- ceph-users@xxxxxxx
To unsubscribe send an email to ceph-users-leave@xxxxxxx
