Upgrading your Ceph cluster software version is relatively simple. You just need to update the Ceph package followed by a service restart and you are done. The upgradation process is sequential and usually does not require any downtime for your storage services if you have configured your cluster in high availability, that is, with multiple OSDs, monitors, MDS, and RADOSGW. As a general practice, you should plan your cluster upgradation during nonpeak hours. The upgradation process upgrades each Ceph daemon one by one. The recommended upgradation sequence for a Ceph cluster is as follows:
- Monitor
- OSD
- Metadata Server (MDS)
- RADOS gateway
You should upgrade all daemons of a specific type and then proceed for upgradation of the next type. For example, if your cluster contains three monitor nodes, 100 OSDS, 2 MDS, and a RADOSGW. You should first upgrade all your monitor nodes one by one, followed by upgrading all OSD nodes one by one, then MDS, and then RADOSGW. This is to keep all specific types of daemons on the same release level.
Proceed with the following steps to upgrade a monitor:
- Check the current version of your Ceph daemons:
Since our test cluster setup has MON and OSD daemons running on the same machine, upgrading Ceph software binaries to the Firefly release (0.80) will result in upgrading MON and OSD daemons in one step. However, in the production deployment of Ceph, upgradation should take place one by one. Otherwise, you might face problems.
- Upgrade your Ceph repositories to the targeted Firefly repositories. Usually, you just need to update your new Ceph release name in the
/etc/yum.repos.d/ceph.repofile that is already present: - Update your Ceph software:
# yum update ceph - Once the Ceph software binaries are updated, you should restart the Ceph monitor daemons to have the changes take effect:
# service ceph restart mon - Check the monitor daemon version; it should be updated:
# service ceph status mon - Check the monitor stats:
# ceph mon stat - Repeat these steps for other monitor nodes in your cluster. Make sure all the monitor nodes are on the same Ceph release before proceeding to OSD, MDS, or RGW upgradation.
Proceed with the following steps to upgrade OSDs:
- Similar to the monitor upgradation process, upgrade your Ceph repositories to the targeted Firefly repositories.
- Perform the Ceph software upgradation:
# yum update ceph - Once the Ceph software is successfully updated, restart the OSD daemons:
# service ceph restart osd - Check the OSD status; you should discover the updated OSD version:
# service ceph status osd - Repeat these steps for other OSD nodes in your cluster. Make sure all OSDs of your cluster are on the same Ceph release.