commit 0d0a165d3665487227432b4b28a2eb819c5ce9aa Author: Patrick McLean <patrick.mclean@sony.com> Date: Wed Jan 20 14:09:39 2021 -0800 sys-cluster/ceph: Revbumps, force older rocksdb (bug #764221) That was >2 years ago. Is it still accurate? Is upstream aware? Is there a bug we can track? Upper bound version limits are bad when there is no bug to check with. Please note that limiting dependencies to a maximum version is merely a workaround, not a bugfix.
Upstream is still bundling rocksdb-6.15.5, I don't have any way to test rocksdb-7 (the problem is a runtime crash when the OSD daemon is started in a cluster, I don't have any cluster that I can test this right now).
Reef claims to support RocksDB 7.9.2 upstream: https://docs.ceph.com/en/reef/releases/reef/#rados I have just tested the RocksDB 7.9.2 + Ceph 18.2.0 combination and it works. To be specific the original bug #764221 prevented startup of the OSDs as far as I can tell, this is no longer the case. The OSDs start up just fine. Whether or not there are any mid- to long-term runtime errors I cannot tell just yet, however the cluster does seem happy with the new RocksDB version.
Short update; going from `<dev-libs/rocksdb-6.15:=` to `<dev-libs/rocksdb-7.10:=` in the ebuild has been running perfectly fine for me for over two weeks now:
see also bug 920807
Yes please.
It appears that bug 892275 got undercut by bug 921204 so now ppc64 is blocking EAPI7Removal...