Re: [HACKERS] Open 6.5 items
От | Tom Lane |
---|---|
Тема | Re: [HACKERS] Open 6.5 items |
Дата | |
Msg-id | 8928.928162998@sss.pgh.pa.us обсуждение исходный текст |
Ответ на | Re: [HACKERS] Open 6.5 items (Vadim Mikheev <vadim@krs.ru>) |
Ответы |
Re: [HACKERS] Open 6.5 items
|
Список | pgsql-hackers |
Vadim Mikheev <vadim@krs.ru> writes: >> As for my test case,xidHash is filled with XactLockTable entries which have >> been acquired by XactLockTableWait(). >> Could those entries be released immediately after they are acquired ? > Ops. Thanks! Must be released. Does this account for the "ShmemAlloc: out of memory" errors we've been seeing? I spent a good deal of time yesterday grovelling through all the calls to ShmemAlloc, and concluded that (unless there is a memory stomp somewhere) it has to be caused by one of the shared hashtables growing well beyond its size estimate. I did find that the PROC structures are not counted in the initial sizing of the shared memory block. This is no problem at the default limit of 32 backends, but could get to be an issue for hundreds of backends. I will add that item to the size estimate today. regards, tom lane
В списке pgsql-hackers по дате отправления: