November 17, 2015 at 8:36 am
Ok sitting at desk Idera sends me alert that it can't connect to cluster. Login and yet it has restarted. No real warnings in log before or after. Only thing my PLE went to basically zero right about the time server crashed. has anyone ever seen PLE problem take down a server?
November 17, 2015 at 8:46 am
The question you probably need to be asking, is what was running that would cause the PLE to go to 0. If you can answer that (and it'll probably need you to have had monitoring on the server already), it may point you at the cause of the crash.
Gail Shaw
Microsoft Certified Master: SQL Server, MVP, M.Sc (Comp Sci)
SQL In The Wild: Discussions on DB performance with occasional diversions into recoverability
November 17, 2015 at 8:55 am
problem is this is a 3rd party app/db with the design code from hell. Idera is good about tracking and takes screen shots when alerts were fired. My PLE has been below 2000 for a while and worse lately. I can run blitzcache and Pinal Dave's script so I can identify the awful that I really can't change. My real question is can a server with 600+ users survive with zero PLE or can that take the server down
November 17, 2015 at 8:57 am
Sure it can. I've seen servers where PLE never made it above 5. Performance sucked, IO subsystem was maxed out, but SQL's not going to crash just because it's reading everything from disk for every request.
Gail Shaw
Microsoft Certified Master: SQL Server, MVP, M.Sc (Comp Sci)
SQL In The Wild: Discussions on DB performance with occasional diversions into recoverability
November 17, 2015 at 9:24 am
and now one of the network guys owned up and made a NIC change that caused the outage. Does not help me with PLE that sucks. I believe this 3rd party app company had their DBA's buried in a land fill or they just jumped out the window. If you wanted to write a whitepaper on how not to design and application and database this is the perfect story.
November 18, 2015 at 6:56 am
But PLE is not a measure of performance. It's merely an indication of load and behavior. Having low or wildly variable PLE is not, by itself an indication of a problem. You need to look at other counters, especially around wait states to get a good understanding of what is going wrong on the system.
"The credit belongs to the man who is actually in the arena, whose face is marred by dust and sweat and blood"
- Theodore Roosevelt
Author of:
SQL Server Execution Plans
SQL Server Query Performance Tuning
November 18, 2015 at 7:05 am
mine is not wildly variable it is bad, 128 gig of RAM on server PLE below 500. Running Ozars Blitzcache and Pinal Dave's cache query I know where my offenders are. Problem is the things that are killing me at 3rd party sql that are part of their interfaces so I can't change. When you do a select * on a 500K row table it is bad, when that table has 6 image data type columns and it does a left outer join to a million row table it is brutal, and this sql is run over a 50 times a minute. 3rd party apps are always a joy.
November 18, 2015 at 7:08 am
Means your IO subsystem will be getting driven very hard, high utilisation, and probably slow queries as a result. As I mentioned, I've seen a server where the PLE never got above 5 in business hours. Slow as hell, but it was still stable.
Gail Shaw
Microsoft Certified Master: SQL Server, MVP, M.Sc (Comp Sci)
SQL In The Wild: Discussions on DB performance with occasional diversions into recoverability
November 18, 2015 at 7:12 am
trust me I have made them aware of this, feel like a lookout on the titantic. getting new server in January and 2012, so if I can limp along till then will make it
Viewing 9 posts - 1 through 8 (of 8 total)
You must be logged in to reply to this topic. Login to reply