On Wed, Dec 14, 2016 at 1:14 AM, Ashutosh Sharma <ashu.coe...@gmail.com> wrote: > I have been working on this issue for last few days and it seems like > i have almost found the reason for this failure. My findings till date > are as follows. > > Just to confirm if the problem is happening due to reusability of > WaitEventSet structure, I had replaced a function call > ModifyWaitEvent() and WaitEventSetWait() which makes use of an already > existing WaitEventSet to store event handles with WaitLatchOrSocket() > and it worked. Actually WaitLatchOrSocket() creates a new WaitEventSet > structure every time it has to wait for an event and is also being > used in the old code. This clearly shows that we do have some problem > just because we are reusing the same set of object handles throughput > a backend session. I am further investigating on this and would post > the final analysis along with the fix very soon. Attached is the > patch that has the changes described > above. Any suggestions or inputs would be appreciated.
So it would mean that there is a mismatch between what should be used for the wait event and what is actually used. One simple idea if you want to compare both would be to generate some elog(LOG) entries or whatever on the fields we are interesting on and see what is mismatching. That may help in understanding what's wrong in the wait events used. -- Michael -- Sent via pgsql-hackers mailing list (firstname.lastname@example.org) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-hackers