Please refer to the following document (page 176):
Recommend before the upgrade:
1) Check the release notes for supported upgrade path, special notices, product integration, known issues and limitations if any.
2) Backup configuration before and after each upgrade.
3) Plan a maintenance window for the upgrade.
3) Have some one available on the remote site in case something went wrong during the upgrade.
4) Make sure check sum is matching between cluster members using the following CLI commands:
# get sys ha status
# diag sys ha showcsum
# execute ha manage <Slave ID> <<-- could be 0 or 1, check "get sys ha status" results
$ diag sys ha showcsum
The results of "diag sys ha showcsum" should be the same on all levels (all/global/vdoms)