Post reply

How reboot of two node of cluster server

  • October 7, 2003 at 8:10 am

    #159934

    Someone can answer this question, we have cluster server which means we can have 24/7 all up, and if we need sometimes just quick reboot, can we just reboot one node then it failover to other node, then we reboot other node it failover to this node again, and we have our server all up running, also how about if we need apply sql server patch can we do same?

    Thanks

    feifei liu


    Rachel

  • October 7, 2003 at 8:23 am

    #477387

    You can use Cluster administrator to fail-over from one node to another node, reboot the first node, fail-over to first node and reboot second node.

    Depending on which patch you are going to apply, in th old time, you have to copy and paste the files from the patch to each SQL Server directory on both node, so you can do it by fail-over as mentioned above to keep SQL Server running.

    But all new patch since ms02-061 ( can't remember exact time) are self installed and cluster aware, which will have to shutdown SQL Server instance you are going to apply and install the files from patch to both nodes at same time. In such installation, SQL Server will be temperorary not available.

  • October 7, 2003 at 8:32 am

    #477393

    Thank you Allen, normally reboot only 3 to 4 minutes, but how long should failover last, from my reading should be around 10 seconds? But someone told me need 3 to 5 minutes? Is That true?

    Thanks

    feifei liu


    Rachel

  • October 7, 2003 at 8:35 am

    #477395

    Definitely it takes less time than reboot. In my SQL Server clusters, it takes about 30 seconds depending on how many resources SQL server resource depend on.

  • October 7, 2003 at 8:46 am

    #477402

    Thank you so much!!!!!

    You answered all my questions!!

    feifei liu


    Rachel

  • October 7, 2003 at 12:11 pm

    #477437

    Failover involves two processes: shutting down one server and starting another. Shutdown can be fast (especially if unplanned!). Starting the new instance depends on how many databases you have and how they're recovered. For example: a server with a database with lots of uncompleted transactions when it fails (could be 1000's or even 10,000's in some environments) may take a substantial time to start up because it goes through a rollback process for each transaction that didn't complete. A manual failover will be slower to shut down and faster to restart than a true failure because the database will complete the pending transactions before it shuts down on the active node.

Viewing 6 posts - 1 through 5 (of 5 total)

You must be logged in to reply to this topic. Login to reply