When upgrading a Symantec Encryption Management Server (PGP Server) as a single node, downtime may be needed. To avoid this, having cluster nodes that manage the same services will help reduce this likelihood.
When upgrading a PGP Server cluster, it is possible to avoid downtime. For example, with a two-member cluster comprising member1 and member2:
If Web Email Protection or PDF Email Protection with an option such as secure replies is enabled, downtime is generally recommended. Consider what happens when the above steps are used to upgrade a two member cluster running Web Email Protection:
As can be seen, at the point where all cluster members have been upgraded, there may be significant differences in the Web Email Protection data between cluster members. The earliest that full synchronization can be expected is 3am on the day following the upgrade but several days may be required for full synchronization.
If something goes wrong during the upgrade or if replication does not work properly after the upgrade, there is a risk that Web Email Protection data will be inconsistent between cluster members for an extended time and in extreme cases there is a risk of complete data loss. These are risks that most organizations are unwilling to take.
Important Tip for WEP to Reduce the need for downtime: Due to some of the operations during these types of upgrades, it is recommended to enable WEP on all nodes for the PGP Server to be able to operate one node at a time. In particular, each PGP Server can have the WEP service disabled, but this usually means that messages will not reside on each of the nodes, making downtime more likely. If you have the WEP service enabled on all nodes, as part of the migration, you can have WEP handle the service until one node is migrated.
Then point traffic to the migrated node, and then you migrate the previous node. Having the WEP service enable don all nodes also makes replication rings more straightforward.
Symantec Encryption Management Server release 10.5 and above.
To avoid the risks described above, some downtime will be necessary: