3 Replies Latest reply: Mar 14, 2014 7:42 AM by Todd Little-Oracle RSS

    Event support for automatic failover in Tuxedo 12c?

    Per Lindström

      Hi all,

       

      I'm experimenting with the new features for automatic failover that were introduced in Tuxedo 12c, i e the parameters DBBLFAILOVER and SGRPFAILOVER.

       

      It all looks very interesting, but I would have been expecting som kind of system event(s) to be able to tell me that automatic failover is decided to happen and in progress. However, I don't seem to find any "suitable" events in the reference doc (EVENTS(5)). Am I simply missing something, looking in the wrong place, or is there a possible product enhancement hidden here?

       

      Any input on this welcome!

       

      Best regards

      /Per

        • 1. Re: Event support for automatic failover in Tuxedo 12c?
          Todd Little-Oracle

          Hi Per,

           

          Are you saying no events get generated?  I would think you would at least get:

          1. .SysNetworkFailure
          2. .SysMachinePartitioned
          3. .SysMachineState
          4. And numerous server related events

           

          Are you not seeing any of these?  Or would you like specific events to indicate that migration is happening automatically?

           

          Regards,

          Todd Little

          Oracle Tuxedo Chief Architect

          • 2. Re: Event support for automatic failover in Tuxedo 12c?
            Per Lindström

            Todd,

             

            at this early design stage I'm not "seeing" any events at all, I'm not seeing anything except documentation, but I hope to setup some hands-on tests this week :-)

             

            My understanding of these events is that they're all fine if I want to detect machine/network failures and do some migrating on my own using MIB calls or scripts, but if I'm using "declarative fail-over" I'd rather get an event when that whole business has been taken care of, and it's time to do any possibly necessary application-level cleanup. Ideally one event in the T_DOMAIN area (MASTER status migrated) and one for each server group having been automatically migrated.

             

            Otherwise I guess I'd listen for the "failure" events that you mentioned, find out the timeout value(s) regarding automatic failover from T_DOMAIN, delay my processing that much and hope I delay enough but not too much. Seems like a long (and winding, if the Beatles will excuse me) road to go :-).

             

            On browsing the EVENTS(5) documentation again, I found a .SysGroupState event that looks somewhat promising - could that be part of what I'm looking for? But I guess I still have to look for the MIGrating state followed by a new ACTive state, or how would this happen? And would this be able to tell an automatically initiated migration (i e due to HW/NW failure, want to do application level cleanup) from a manually (i e planned maintenance, application cleanup probably not necessary) initiated migration?

             

            On another note - I guess you really have to setup a TMSYSEVT server on all MACHINE:s in order to get anything like this having any kind of reliability... or is the key to keep the TMSYSEVT server running on the same node as the DBBL to get all these events distributed, or will any ole BBL do the actual job of posting them?

             

            Best regards

            /Per

            • 3. Re: Event support for automatic failover in Tuxedo 12c?
              Todd Little-Oracle

              Hi Per,

               

              This is clearly an area of documentation could use a little work. :-(   In any case, I think it is reasonable to have events indicating a machine/group is migrating and one that indicates whether it completed successfully or not.  Care to submit an ER?  :-)

               

              Regards,

              Todd Little

              Oracle Tuxedo Chief Architect