Hello,

We are investigating a problem with unwanted failovers on our VMWare platform during disk conslditation after a Veeam backup. We are running a SQL2012 mirror with a witness server

We realize that the the problem has to do with the disk consolidations however I have one question regarding the mirroring.

We notice that the principal server performs a failover to the mirror server after reaching the failover timeout in the mirroring connection to the mirror and the witness which we set at 60 sec. This is however several seconds before the mirror decides to become principal. Is it correct that a principal server which loses connection to both mirror and witness will change his role to mirror?

Thx

Erik

Hello Erik,

So if you are running in a High Availability mirroring (using witness with automatic failover) the case is that the mirror and the witness decide to switch role the role of the mirror based on the condition they cannot reach the principal. Additional explanation you can find in the Automatic Failover section in this article.

Hello Ivan,

I understand that the mirror and can decide to sitch the role. However, as stated in my question, the first thing that happens is that the pricipal becomes mirror, after that the mirror becomes principal. If the mirror and the witness cannot contact the principal, why does the principal switch role first?

Hence the question:

Can the principal decide on its own to become mirror when he cannot reach both mirror and witness.

Regards

Hello Erik,

Sorry, I seem to have misunderstood your question. 

The principal server cannot decide to become a mirror on his own. The decision point stays with the witness - if he cannot perform his "is alive" check to the principal, he tries to check the mirror and if that goes well, the roles are switched. In all cases, in order roles to switch, you should have a not responding principal (cannot accept new connections for instance, or for some time there were network glitches). 

Now imagine a situation where your principal is not reachable for 65 seconds - the witness and mirror take the decision to switch the roles. The moment your principal is available again, it talks to the witness, notices that the roles have now changed hence he is now a mirror.

Hi MS_Erik,

If the principal server can't reach both mirror server and witness, the principal server couldn't become the mirror server automatically according to lvan's post. That is to say, you couldn't perform automatic failover. If the mirror server and witness are connected to each other in high-performance mode or high-safety with automatic failover mode, you need to perform force failover.

About Role Switching During a Database Mirroring Session, please refer to the article: https://msdn.microsoft.com/en-us/library/ms189850.aspx#TypicalCaseFS

Regards,
Michelle Li

• Edited by Michelle LiMicrosoft contingent staff 22 hours 16 minutes ago

Hi,

Thank you for your answer, it was something I needed to rule out.

Could you eloborate on what the 'alive check' exactly encompasses? The reason for asking is we have extended the timeout to 180 seconds with no succes. At the same time we have a ping and port 1433/5022 check every 30 seconds that shows no downtime!

Regards,

Erik