Thursday, July 19, 2012
been awhile since last blogged...good to be back and jog down whatever in my head... and as usual i not care much about spelling or grammar... it's been a very challenging experiences. I definitely learn a lot. I'd done clustering back in windows 2003 and sql server 2005. With win2008R2/sql2012 combination, oh lots of changes... all in good way. We only have 2-node cluster and will be active/active config, but we almost ran out of drive letters. Then we decided to use mounted volumnes rather than drive letter. Tempdb will be on shared disk as well as other cluster volumnes. One improvement in sql 2012 is now you can put tempdb on local disk if you wish. This helps a lot in performance when you have tempdb on some sort of local faster drives (eg. SSD). I have to admit, in clustering world, we have many ip addresses along with associated names. To name a few: - physical node name and their ip - windows cluster name and ip - sql server cluster virtual name and ip (1 pair for 1 sql instance) - sql server alias for app and ip (1 pair for 1 sql instance) - cluster msdtc and ip (1 pair for each sql instance) I have 2-node with active/active: look like i got 9 "name" and 9 ip these first 2-node cluster will be our OLTP. Once this up and running, my next project would set up 2nd 2-node cluster for some sort of OLAP ... kind of late... be back next time.