Ceph Octopus

[kyriazis]

Looking at the upgrade instructions, I have some questions:
- Do i have to upgrade packages on all nodes at the same time, or can I do the upgrade gradually? For example, update packages on hosts with mgr daemons first, then osds, etc.
- The instructions make no reference to reboots at any stage of the process. Is that accurate? Maybe one will be needed after the initial full-upgrade?

Thanks!
George

 

[spirit]

Looking at the upgrade instructions, I have some questions:
- Do i have to upgrade packages on all nodes at the same time, or can I do the upgrade gradually? For example, update packages on hosts with mgr daemons first, then osds, etc.
- The instructions make no reference to reboots at any stage of the process. Is that accurate? Maybe one will be needed after the initial full-upgrade?

Thanks!
George

upgrade package on all nodes (apt dist-upgrade),

then restart ceph-mon service on each node, 1 by 1 (systemctl restart ceph-mon.target) . wait for ceph health status ok between each restart
then restart ceph-osd sernice on each node 1 by 1 (systemctl restart ceph-osd.target) . wait for ceph health status ok between each restart
the restart ceph-mds service on each node 1 by 1 (systemctl restart ceph-mds.target). wait for ceph health status ok between each restart

done

 

[kyriazis]

Documentation says "apt full-upgrade" not "apt dist-upgrade". Any difference?

 

[t.lamprecht]

Documentation says "apt full-upgrade" not "apt dist-upgrade". Any difference?

no, those are aliases.

 

[kyriazis]

I can confirm that I was able to upgrade a 24-node proxmox cluster (17 ceph nodes; 39 osds) from nautilus to octopus with no problems. ceph was in service during the upgrade, and no problems occurred.

Thank you!