I'm on quincy.
I had lots of problems with RGW getting stuck. Once I dedicated 1
single RGW on each side to do replication, my problems went away.
Having a cluster of RGW behind a load balancer seemed to be confusing
things.
I still have multiple RGW for user-facing load, but a single RGW
So we still have this rgw synchronization that gets stuck every day and
about the same time. We have alerting on it, so our on-call people are
getting annoyed.
Summarizing: we see on the receiving end of an rgw sync something like this:
# radosgw-admin sync status --rgw-realm backup
On 08.08.24 08:31, Eugen Block wrote:
>> Redeploying stuff seems like a much too big hammer to get things
going again. Surely there must be something more reasonable?
>
> wouldn't a restart suffice?
Probably, but when we were handling this the first time around, a
redeploy was the first thing
Hi,
Redeploying stuff seems like a much too big hammer to get things
going again. Surely there must be something more reasonable?
wouldn't a restart suffice?
Do you see anything in the 'radosgw-admin sync error list'? Maybe an
error prevents the sync from continuing?
Zitat von Olaf Seibe
