RSS Subscription 168 Posts and 2,769 Comments

CCR Unscheduled Failovers and Database Mounting Issues

I’ve seen several people on forums talking about CCR unscheduled failovers not failing over correctly and databases not being mounted properly. I did quite a bit of testing today on this and found a couple things. First thing I noticed is that with no Rollups installed and a default CCR configuration (Best Availability) that when you do an unscheduled failover, such as pulling the network cable, the databases failover to the passive node but the storage groups sit at online pending and then go to failed. When you bring the old active server back, replication ensues and the storage groups go from failed to online. For those that work on CCR clusters should know that there are 3 CCR availability settings to choose from.

  1. Lossless
  2. Good Availability
  3. Best Availability (Default)

The above situation is typical of Lossless. When a Lossless failover occurs, the databases most likely won’t mount. This is because with lossless, the databases/logs have to be in perfect condition. In order for the passive node (new active) to be able to mount the databases, the administrator has to intervene or you will have to get the old active up and running so replication can resume which will result in the database being able to mount.

Good availability and Best Availability are very similar in that they will allow the databases to be mounted at the cost of losing 3 (good) or 6 (best) availability. Of course with this, you will have to accept that you “might” lose some data if your Transport Dumpster configuration settings are improperly configured. When you bring back up the old active which will be the new passive, to make sure that databases/logs are consistent, you’ll want to do a re-seed.

Now with the issue I explained in the first paragraph, best availability is not working as intended. If you are seeing this issue, make sure you install the latest Rollup which is currently at Rollup 4. Go to my other article here for more information in regards to Rollup 4. As soon as I installed Rollup 4, this issue started to go away and whenever an unscheduled failover occured, it would failover properly and the storage groups would come online and the databases would mount.

Share

Trackback this post | Feed on Comments to this post

Leave a Reply