Обсуждение: Fatal process interaction

Поиск
Список
Период
Сортировка

Fatal process interaction

От
"Steven M. Wheeler"
Дата:
The situation:
In xterm #1: psql running, connected to database pssdb, executing an SQL select
statement.

In xterm #2: A shell script is run that pipes another select statement throug psql.

When the select statement in xterm #2 completes, the select statement being executed
in xterm #1 aborts and the following error message is output:

NOTICE:  Message from PostgreSQL backend:       The Postmaster has informed me that some other backend died abnormally
and
possibly corrupted shared memory.       I have rolled back the current transaction and am going to terminate your
database system connection and exit.       Please reconnect to the database system and repeat your query.
pqReadData() -- backend closed the channel unexpectedly.       This probably means the backend terminated abnormally
beforeor while
 
processing the request.
We have lost the connection to the backend, so further processing is impossible.
Terminating.

This problem is very repeatable and does not seem to depend on which UID is running
which session.  They can both be postgres, or me (smw) or one of each.  The one
common factor appears to be that psql is involved in both.  If psql is run in
interactive mode in both windows, then \q is issued in one of the windows, both psql
sessions will terminate.

Any ideas?
--
Steven M. Wheeler
UNIX Engineering
The SABRE Group
(918) 292-4119
(918) 292-4165 FAX




Re: [SQL] Fatal process interaction

От
Tom Lane
Дата:
"Steven M. Wheeler" <swheeler@sabre.com> writes:
> If psql is run in interactive mode in both windows, then \q is issued
> in one of the windows, both psql sessions will terminate.

Wow, that's pretty bizarre.  What platform are you on?  I think the FAQ
mentions that problems like this have been seen on systems where SysV
semaphore support wasn't configured or didn't work right ...
        regards, tom lane


Re: [SQL] Fatal process interaction

От
"Steven M. Wheeler"
Дата:
More info.  Motivated by your remark about SysV semaphores, I upgraded the
kernel to 2.0.36 and the problem appears to have gone away.
--
Steven M. Wheeler
UNIX Engineering
The SABRE Group
(918) 292-4119
(918) 292-4165 FAX

Tom Lane wrote:

> "Steven M. Wheeler" <swheeler@sabre.com> writes:
> > If psql is run in interactive mode in both windows, then \q is issued
> > in one of the windows, both psql sessions will terminate.
>
> Wow, that's pretty bizarre.  What platform are you on?  I think the FAQ
> mentions that problems like this have been seen on systems where SysV
> semaphore support wasn't configured or didn't work right ...
>
>                         regards, tom lane



Re: [SQL] Fatal process interaction

От
Tom Lane
Дата:
"Steven M. Wheeler" <swheeler@sabre.com> writes:
> More info.  Motivated by your remark about SysV semaphores, I upgraded the
> kernel to 2.0.36 and the problem appears to have gone away.

You know, you're the second person in the past few months who has
reported strange problems that stopped after a kernel upgrade.
Which Linux version were you on before?  This definitely needs to
go in the FAQ, I think...
        regards, tom lane


Re: [SQL] Fatal process interaction

От
"Steven M. Wheeler"
Дата:
This system was originally configured from Slackware 3.6, kernel 2.0.35.

Tom Lane wrote:

> "Steven M. Wheeler" <swheeler@sabre.com> writes:
> > More info.  Motivated by your remark about SysV semaphores, I upgraded the
> > kernel to 2.0.36 and the problem appears to have gone away.
>
> You know, you're the second person in the past few months who has
> reported strange problems that stopped after a kernel upgrade.
> Which Linux version were you on before?  This definitely needs to
> go in the FAQ, I think...
>
>                         regards, tom lane

--
Steven M. Wheeler
UNIX Engineering
The SABRE Group
(918) 292-4119
(918) 292-4165 FAX