Click here to monitor SSC
SQLServerCentral is supported by Redgate
Log in  ::  Register  ::  Not logged in
Home       Members    Calendar    Who's On

Add to briefcase

Slowness on a cluster. Expand / Collapse
Posted Wednesday, August 28, 2013 7:44 AM
SSC Journeyman

SSC JourneymanSSC JourneymanSSC JourneymanSSC JourneymanSSC JourneymanSSC JourneymanSSC JourneymanSSC Journeyman

Group: General Forum Members
Last Login: Tuesday, November 12, 2013 9:40 AM
Points: 81, Visits: 29
Good morning all,

I am not sure whether this belongs in this forum.

Let me preface this by saying I do not have a huge amount of experience with clusters.
We have an active/active cluster. One side has the default instance hosting 97 databases.
The other side has three instances each supporting a single application. (Note: I did not set this up).
There is 64 GB of ram and 16 processors on each one.
We are running SQL Server 2008 R2 SP2 CU5.

We are getting slowness complaints from users of one application whose databases are in the default instance. There are 20 separate databases supporting this app. The slowness comes and goes. I ran a trace a few weeks ago that seemed to show that all applications were affected during these times.
If I look at resource waits, I see issues with Logging and Buffer I/O have the highest cumulative wait times. Backup is also in the same ballpark. (2,000,000)+.
When I see a wait type pop up under processes, it is usually WRITE_LOG.
There are seven 4GB temp files, but all on the same drive. We have added three addition drives to the default instance and plan to move 2 of the temp files to each one.
That would leave 1 temp file and the temp log on the original drive. I know how to reassign the files. My question is, will restarting the service cause it to fail over to the B side?
Am I even on the right track here?
I have included top 20 wait times.

  Post Attachments 
Top 20 wait times.xlsx (3 views, 10.49 KB)
Post #1489220
Posted Wednesday, August 28, 2013 3:15 PM

SSC Eights!

SSC Eights!SSC Eights!SSC Eights!SSC Eights!SSC Eights!SSC Eights!SSC Eights!SSC Eights!

Group: General Forum Members
Last Login: Today @ 3:29 AM
Points: 889, Visits: 862
Your highest wait type is CXPACKET, which indicates that you have a lot of (unbalanced) parallelism.

My approach to a situation like this is to set up a trace where I capture Stmt:Completed and a few more events filtered for Duration, for instance > 100 ms. But that is more difficult if the application does not use stored procdures, and even more if they don't parameterise statements. But apart from that, this usually gives me some queries to tune (by rewriting, adding indexes etc.)

And since this problem comes and goes this indicates that there is a certain process that needs fixing.

Erland Sommarskog, SQL Server MVP,
Post #1489427
« Prev Topic | Next Topic »

Add to briefcase

Permissions Expand / Collapse