1 Upgrading the MDS Cluster
2 =========================
4 Currently the MDS cluster does not have built-in versioning or file system
5 flags to support seamless upgrades of the MDSs without potentially causing
6 assertions or other faults due to incompatible messages or other functional
7 differences. For this reason, it's necessary during any cluster upgrade to
8 reduce the number of active MDS for a file system to one first so that two
9 active MDS do not communicate with different versions. Further, it's also
10 necessary to take standbys offline as any new CompatSet flags will propagate
11 via the MDSMap to all MDS and cause older MDS to suicide.
13 The proper sequence for upgrading the MDS cluster is:
15 1. Disable and stop standby-replay daemons.
19 ceph fs set <fs_name> allow_standby_replay false
21 In Pacific, the standby-replay daemons are stopped for you after running this
22 command. Older versions of Ceph require you to stop these daemons manually.
26 ceph fs dump # find standby-replay daemons
30 2. Reduce the number of ranks to 1:
34 ceph fs set <fs_name> max_mds 1
36 3. Wait for cluster to stop non-zero ranks where only rank 0 is active and the rest are standbys.
40 ceph status # wait for MDS to finish stopping
42 4. Take all standbys offline, e.g. using systemctl:
46 systemctl stop ceph-mds.target
48 5. Confirm only one MDS is online and is rank 0 for your FS:
54 6. Upgrade the single active MDS, e.g. using systemctl:
58 # use package manager to update cluster
59 systemctl restart ceph-mds.target
61 7. Upgrade/start the standby daemons.
65 # use package manager to update cluster
66 systemctl restart ceph-mds.target
68 8. Restore the previous max_mds for your cluster:
72 ceph fs set <fs_name> max_mds <old_max_mds>
74 9. Restore setting for ``allow_standby_replay`` (if applicable):
78 ceph fs set <fs_name> allow_standby_replay true
81 Upgrading pre-Firefly file systems past Jewel
82 =============================================
86 This advice only applies to users with file systems
87 created using versions of Ceph older than *Firefly* (0.80).
88 Users creating new file systems may disregard this advice.
90 Pre-firefly versions of Ceph used a now-deprecated format
91 for storing CephFS directory objects, called TMAPs. Support
92 for reading these in RADOS will be removed after the Jewel
93 release of Ceph, so for upgrading CephFS users it is important
94 to ensure that any old directory objects have been converted.
96 After installing Jewel on all your MDS and OSD servers, and restarting
97 the services, run the following command:
101 cephfs-data-scan tmap_upgrade <metadata pool name>
103 This only needs to be run once, and it is not necessary to
104 stop any other services while it runs. The command may take some
105 time to execute, as it iterates overall objects in your metadata
106 pool. It is safe to continue using your file system as normal while
107 it executes. If the command aborts for any reason, it is safe
108 to simply run it again.
110 If you are upgrading a pre-Firefly CephFS file system to a newer Ceph version
111 than Jewel, you must first upgrade to Jewel and run the ``tmap_upgrade``
112 command before completing your upgrade to the latest version.