HCL Notes and Domino 9.0 Social Edition Forum - Resource not found in schedule database

Oct 3, 2016, 4:57 PM

34 Posts

A quick followup

The clustering failover design was added R7 to make R&R more robust in the event of a server failure. Based on our several factors we limited the failover mechanism to happen on between the R&R dBs primary server and the next clustermate when the list of clustermates is sorted alphabetically. That can sound a little confusing so let me demonstrate using a cluster of 3 server: Server A, Server B, and Server C.

If the primary server is Server A then failover will happen between Server A and Server B, never Server C. This is also the case if the primary server is Server B. That is because when we sort the list of clustermates and remove the primary server Server B (or Server A in the 2nd example) come before Server C. In order to have Server C be involved with failover it would need to be the primary server. In that case failover would between Server C and Server A but never Server B.

So if you set mail1 to be the primary server then failover should only happen to mail2. It should never happen with mail3 unless mail3 is set as the primary server. The only way I can see for mail3 to attempt failover and take control when mail1 died was if mail3 did not know about mail2.

You can see when the last time any server took control of an R&R database if you set DEBUG_RNRMGR=64 and then do a whoowns command. The INI will cause RnRMgr to generate some extra output that tells you when each server in the cluster last attempted failover for the database you specify. For example:

tell rnrmgr whoowns Rooms.nsf
CheckDBOwnership> NoteID NT00000C9E indicates replica 852563E3:007A46DF failedover on 09/26/2016 11:26:18 AM to server Server A/Kahn
CheckDBOwnership> NoteID NT00000A2A indicates replica 852563E3:007A46DF failedover on 08/08/2016 07:29:54 AM to server Server B/Kahn
Database IrisConf.nsf is currently under the control of server Server A/Kahn (0)

The CheckDBOwnership> tagged lines tell me that Server B had control for most of August and September and that Server A only recently took over processing of requests in my test dB.

Once you have control back on mail1 I would check to see if failover happens with mail2 or mail3 in the future. If you see it failover to mail3 then please contact Technical Support so we can investigate why and work on a solution to keep it off your DR server.

In case you were not aware, the 'primary' server is specified in the R&R database ACL. It is the Administration Server setting. It is used to set the Home Server setting of each room doc in the NAB.

Bruce
IBM