Thread: Lock up, but not deadlock?

Lock up, but not deadlock?

From
Daniel Åkerud
Date:
Is there a possibility that there will be a lock in a transaction, but it is not a deadlock?
I have N functions (about 20-25) interacting in a system. They are like: setUserOnline(...) setUserOffline(...) setServerOnline(...) and so forth. We have found out that there are sometimes deadlocks. We have solved this by trying again 5 times. Most oftenly it works after the first retry, and worst case has been 2 so far. Anyway, is there a possibility that these functions (we have NO manual lock obtaining in them) cause lockups that is not called deadlocks but has to be solved anyway? Like an UPDATE waiting for another statement in another function, but it's not a circular thingy (a.k.a deadlock). Thanks for any help!
 
Daniel Åkerud

Re: Lock up, but not deadlock?

From
Tom Lane
Date:
=?iso-8859-1?Q?Daniel_=C5kerud?= <zilch@home.se> writes:
> Is there a possibility that there will be a lock in a transaction, but it i=
> s not a deadlock?
> I have N functions (about 20-25) interacting in a system. They are like: se=
> tUserOnline(...) setUserOffline(...) setServerOnline(...) and so forth. We =
> have found out that there are sometimes deadlocks.

What PG version?

There used to be some bugs in the deadlock detector, but there are no
known bugs there as of 7.2.

            regards, tom lane

Re: Lock up, but not deadlock?

From
Daniel Åkerud
Date:
We use 7.1.3,
an upgrade is due next thing in the morning :-)

thanks.

----- Original Message -----
From: "Tom Lane" <tgl@sss.pgh.pa.us>
To: "Daniel Åkerud" <zilch@home.se>
Cc: <pgsql-general@postgresql.org>
Sent: Thursday, March 07, 2002 9:29 PM
Subject: Re: [GENERAL] Lock up, but not deadlock?


> =?iso-8859-1?Q?Daniel_=C5kerud?= <zilch@home.se> writes:
> > Is there a possibility that there will be a lock in a transaction, but
it i=
> > s not a deadlock?
> > I have N functions (about 20-25) interacting in a system. They are like:
se=
> > tUserOnline(...) setUserOffline(...) setServerOnline(...) and so forth.
We =
> > have found out that there are sometimes deadlocks.
>
> What PG version?
>
> There used to be some bugs in the deadlock detector, but there are no
> known bugs there as of 7.2.
>
> regards, tom lane
>