New evidence! We've got a new repro of a similar (albeit not identical) case, and I've had the opportunity to dig in deeper and get better data. So far it looks like the problem is actually inside ADO.NET -- the SqlDataReader is waiting in COM code -- on an unmanaged lock -- and the wait times out even though the server does send a response. However, executing the same query using raw ADO.NET code (even using a SqlConnection created by the NHibernate session!) still returns almost instantly, so the best hypothesis I've come up with is that NHibernate configures the command or the reader in a way that triggers the problem. I'll try to investigate further on Monday if the problem still persists and if I manage to set up both .NET and NHibernate debugging symbols.
-Lauri On Monday, October 20, 2014 1:05:08 PM UTC+3, Lauri Kotilainen wrote: > > > > On Friday, October 10, 2014 7:33:57 AM UTC+3, Darren Kopp wrote: >> >> I would guess that the performance you see on your machine is the same as >> on Azure, but you're likely hitting your cpu limits. I'm not sure ordering >> would have any issue on hydration, unless perhaps you have a poor hash code >> algorithm and perhaps were hitting worst case scenarios on hash tables, but >> that also seems pretty suspect. Do you have any .net profiling data >> comparing the two scenarios you outlined? >> > > I did, but I didn't save it -- and guess what, by the end of the day, the > probelm disappeared by itself. > > The hot spots found by the profiler were all somewhere in NHibernate's > proxy generation code paths, which seemed extremely weird to me. > > I'm not sure how to interpret your comment on CPU limits. Maybe I wasn't > very clear in my original post: running the query on the same computer (my > dev machine) produced wildly different results with the same code, > depending on which database I was hitting -- even though my local database > was a very recent (minutes old) copy of the production DB, so the data sets > were nearly identical. And removing the OrderBy eliminated the difference. > > I'll get back to this issue if it reappears and I manage to gather some > more diagnostic data. > > -Lauri > -- You received this message because you are subscribed to the Google Groups "nhusers" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To post to this group, send email to [email protected]. Visit this group at http://groups.google.com/group/nhusers. For more options, visit https://groups.google.com/d/optout.
