Greetings. 2008 R2, SP1.
We've got a sproc that runs all the time with sub second response. All of a sudden this morning the query started timing out at 30 seconds. We got the execution plan saved off, did a recompile, and our sub second response times were back. The new execution plan is much prettier. The main difference (though there are many differences) between the two is that the bad one is full of parallelism, but none of them are showing any actual cost, so I'm not wanting to hang my hat on that.
So how would this suddenly occur? Page Life Expectancy was very low just before this took effect, which I know would slow down the query, but I don't think that could make it actually become a bad execution plan? It should also be noted that this server does run into low PLE's from time to time without this occurring.
No sudden massive increase in data. Stats were updated about 4 hours before this occurred. So how could this have suddenly have occurred? How can I reproduce it to make a query tuning recommendation?
TIA, ChrisRDBA