Hello,

We have got a cephadm-managed test Ceph cluster created with a Tentacle 
Release Candidate, RC1 if memory serves me right.

With 20.2.0 having officially been releases mid-November, we would 
ideally like to bring this cluster into production without data loss - 
none of the data already present on the OSDs is irreplaceable but we 
would prefer to retain snapshot history we have established during the 
test phase.

Unfortunately, it turns out that the naive approach of having "ceph orch 
upgrade" upgrade the cluster to a production release does not work - if 
we run it with "--ceph-version 20.2.0" it immediately fails with the 
error "cannot downgrade to a dev version", "--image 
quay.io/ceph/ceph:v20.2.0" on the other hand appears to start 
successfully but gets immediately paused, with the same error message 
appearing in the upgrade status report.

The "downgrade" bit is likely related to the fact that according to 
"ceph versions", the Ceph version we have presently got running is 
20.3.0-2957-g62bcf65e (despite the fact the cephadm package we used 
clearly having been labelled 20.1.something). The "to a dev version" 
bit, however, has got me stumped given the long-established Ceph has 
been using the "even minor versions are production" convention for ages.

Any thoughts on how we can get this cluster upgraded to a production 
release without wiping everything and starting from scratch? Thanks in 
advance!

PS. cephadm on the admin node has already been upgraded to 20.2.0.

-- 
MS

_______________________________________________
ceph-users mailing list -- [email protected]
To unsubscribe send an email to [email protected]

Reply via email to