Skip site navigation (1) Skip section navigation (2)

Re: some other backend died abnormally

From: fredrik chabot <fredrik(at)f6(dot)nl>
To: Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us>
Cc: Joe Conway <mail(at)joeconway(dot)com>, pgsql-bugs(at)postgresql(dot)org
Subject: Re: some other backend died abnormally
Date: 2002-09-24 19:38:24
Message-ID: (view raw, whole thread or download thread mbox)
Lists: pgsql-bugs
Tom Lane wrote:

>fredrik chabot <fredrik(at)f6(dot)nl> writes:
>>so I migrated the database to a non-smp machine running RH 7.2 and the postgres
>>7.2.1-5 rpm's
>>Kernel 2.4.7-10
>>and I still get
>>FATAL 1:  LWLockAcquire: can't wait without a PROC structure
>>The code tells me this should never heppen :->
>I don't think it should happen either ;-)  When do you see it exactly
>--- is it during startup of a fresh connection, or while executing
>queries in an existing backend, or ...?
Ok, short answer I don't know; long answer:

Most connections are "long-lived" connections staying up for hours on 
end. There are somewhere form 6  upto ~75 connections open at a time 
about 6 at night > 50 during "office hours". Both failures today where 
in office hours and It happend about every hour in office hours on the 
smp system and only once in the off hours.

I got it BTW twice while loading the data in the database. Simple stuff 
begin; -thousands of inserts; commit; and then the next table.
I did not once fail during the test fase of the system. (one or two 
users at a time)

So my hunch is that it happens when there has been a lot of operations 
in one connection and a lot of other connections at the same time.

>			regards, tom lane
>PS: please don't send HTML mail to the lists, it's a pain to quote ...
>---------------------------(end of broadcast)---------------------------
>TIP 2: you can get off all lists at once with the unregister command
>    (send "unregister YourEmailAddressHere" to majordomo(at)postgresql(dot)org)

In response to

pgsql-bugs by date

Next:From: pgsql-bugsDate: 2002-09-26 14:43:31
Subject: Bug #780: Timestamp shifted by one minute
Previous:From: Tom LaneDate: 2002-09-24 16:23:22
Subject: Re: some other backend died abnormally

Privacy Policy | About PostgreSQL
Copyright © 1996-2017 The PostgreSQL Global Development Group