Dear dojo community,
Question about Failover configuration.
I believe it is not possible to mix the two failover methods in the same DMS, correct?
In case some DMA’s in a DMS are configured for Failover using a Virtual IP but there is a need to add a failover pair using a shared hostname. Then the existing failover configuration needs to migrate to a shared hostname as well. What is the best way to migrate an existing failover pair from a virtual IP to a shared hostname? Is it advised to breakup the failover config and then re-create it using a shared hostname?
Thanks!
Interesting question – can I ask why the hostname config would be needed just for one pair of DMAs? Is it an intermediate step or just a new DMA pair that will be landed by a different set of users?
I haven’t tested this myself, but I believe you’re right as the documentation about Failover Config in Cube indeed suggests to always use only one method:
step 8 & 9 in the guide seem mutually exclusive:
Curious if out of CUBE a hybrid set-up could be tested – but to stay within what’s recommended in that section, if the cluster is not too big, I believe each pair would have to be reconfigured – I think this is recommended also for SLNet and SLFailover logging.
I’d check also the DMS.xml of the pairs – the newest Failover configuration should be automatically synchronized between the two DMAs – and DataMiner.xml for the order of NICs:
In my sample I have only “Virtual IP” tags – worth checking how this evolves while transitioning to shared hostname:
HTH
Alberto, the need to use hostname based failover for a specific DMA pair comes from the requirement to have Geo Redundancy for a new DMA to be added to the existing DMS (DMA's in different sites/networks).
In the meanwhile it was confirmed that the recommended way to migrate from VIP to shared hostname is indeed by breaking up the "VIP" failover pair first – double check the backup agent such that is in a good "clean" state and then reconfigure failover using a shared hostname.