Postgres Pro
Downloads
Home   >   mailing lists

Thread: Bad behaviour in Sun Cluster

Bad behaviour in Sun Cluster

From
Patricio Mora
Date:
Hello all.

- postmaster (PostgreSQL) 7.4.5 (Update unviable due to application)
- Solaris 9
- Sun Cluster 3.1.0
- SUNWscPostgreSQL 3.1.0

About 3 times each week, the check_pgs function of the SUNWscPostgreSQL
bin/functios file, stops my Postgres database.
I enabled the cluster's monitor logs (DEBUG=ALL on etc/config) and found
it complains about a time out when manipulating the tables of the
Postgres user used for the cluster to test the health of the DB server.

The command it throws is this:

/usr/bin/su - postgres -c   /usr/bin/env
LD_LIBRARY_PATH=/usr/local/lib:/usr/local/pgsql/lib /usr/bin/env
PGPASSWORD=XXXX /usr/bin/env PGDATA=/postgres/data /usr/bin/env
PGPORT=5432 /usr/local/pgsql/bin/psql -d sc3_test_database -U scuser -c
"truncate sc3_table;insert into sc3_table (sccol) values('hello im
there');select * from sc3_table" >
/tmp/postgres_server-postgres-tbl-out  2>
/tmp/postgres_server-postgres-tbl-err

A minute after the timeout occurs. Postgres.log doesn't show anything (I
dont log SQL instructions)

Someone's seen something similar? TIA.

Re: Bad behaviour in Sun Cluster

From
"Scott Marlowe"
Date:
On Fri, Oct 31, 2008 at 5:00 AM, Patricio Mora
<patricio.mora@juntadeandalucia.es> wrote:
> Hello all.
>
> - postmaster (PostgreSQL) 7.4.5 (Update unviable due to application)

And this prevents you from updating to 7.4.22?

> - Solaris 9
> - Sun Cluster 3.1.0
> - SUNWscPostgreSQL 3.1.0
>
> About 3 times each week, the check_pgs function of the SUNWscPostgreSQL
> bin/functios file, stops my Postgres database.
> I enabled the cluster's monitor logs (DEBUG=ALL on etc/config) and found it
> complains about a time out when manipulating the tables of the Postgres user
> used for the cluster to test the health of the DB server.
>
> The command it throws is this:
>
> /usr/bin/su - postgres -c   /usr/bin/env
> LD_LIBRARY_PATH=/usr/local/lib:/usr/local/pgsql/lib /usr/bin/env
> PGPASSWORD=XXXX /usr/bin/env PGDATA=/postgres/data /usr/bin/env PGPORT=5432
> /usr/local/pgsql/bin/psql -d sc3_test_database -U scuser -c "truncate
> sc3_table;insert into sc3_table (sccol) values('hello im there');select *
> from sc3_table" > /tmp/postgres_server-postgres-tbl-out  2>
> /tmp/postgres_server-postgres-tbl-err
>
> A minute after the timeout occurs. Postgres.log doesn't show anything (I
> dont log SQL instructions)

Are there any errors in the logs?

Re: Bad behaviour in Sun Cluster

From
Patricio Mora
Date:
Scott Marlowe escribió:
<blockquote
 cite="middcc563d10810310603k52b0b63emb3809a6e8123dd59@mail.gmail.com"
 type="cite">
  On Fri, Oct 31, 2008 at 5:00 AM, Patricio Mora
<patricio.mora@juntadeandalucia.es> wrote:


    Hello all.

- postmaster (PostgreSQL) 7.4.5 (Update unviable due to application)



And this prevents you from updating to 7.4.22?


Thanks, I'll try it on my development environment. I tried some 8
version and application crashed.
<blockquote
 cite="middcc563d10810310603k52b0b63emb3809a6e8123dd59@mail.gmail.com"
 type="cite">



    - Solaris 9
- Sun Cluster 3.1.0
- SUNWscPostgreSQL 3.1.0

About 3 times each week, the check_pgs function of the SUNWscPostgreSQL
bin/functios file, stops my Postgres database.
I enabled the cluster's monitor logs (DEBUG=ALL on etc/config) and found it
complains about a time out when manipulating the tables of the Postgres user
used for the cluster to test the health of the DB server.

The command it throws is this:

/usr/bin/su - postgres -c   /usr/bin/env
LD_LIBRARY_PATH=/usr/local/lib:/usr/local/pgsql/lib /usr/bin/env
PGPASSWORD=XXXX /usr/bin/env PGDATA=/postgres/data /usr/bin/env PGPORT=5432
/usr/local/pgsql/bin/psql -d sc3_test_database -U scuser -c "truncate
sc3_table;insert into sc3_table (sccol) values('hello im there');select *
from sc3_table" > /tmp/postgres_server-postgres-tbl-out  2>
/tmp/postgres_server-postgres-tbl-err

A minute after the timeout occurs. Postgres.log doesn't show anything (I
dont log SQL instructions)



Are there any errors in the logs?


Postgres doesn't show any log (I'm logging only connections)
Cluster says that the service is broken.
And SUNWscPostgreSQL says:

15:45:47 The probe command </opt/SUNWscPostgreSQL/bin/control_pgs -R
postgres_server -G postgres_rg -P /postgres/data/postgres.conf.cluster
probe> timed out
15:45:47 Restarting the resource postgres_server.

Re: Bad behaviour in Sun Cluster

From
"Scott Marlowe"
Date:
On Fri, Oct 31, 2008 at 7:17 AM, Patricio Mora
<patricio.mora@juntadeandalucia.es> wrote:
> Scott Marlowe escribió:
>
> On Fri, Oct 31, 2008 at 5:00 AM, Patricio Mora
> <patricio.mora@juntadeandalucia.es> wrote:
> > > - postmaster (PostgreSQL) 7.4.5 (Update unviable due to application)
>
> > And this prevents you from updating to 7.4.22?
>
>
> Thanks, I'll try it on my development environment. I tried some 8 version
> and application crashed.

Yeah, there's lots of changes from 7.4.x to 8.0 and up.  But, 7.4.22
should be compatible with 7.4.5.  There's a lot of bug fixes, some
rather nasty ones, between 7.4.5 and 7.4.22.

> > > Are there any errors in the logs?
>
>
> > Postgres doesn't show any log (I'm logging only connections)
> > Cluster says that the service is broken.
> > And SUNWscPostgreSQL says:

Well, without any more logging you're not likely to get anymore useful
info.  I'd decrease the logging threshold to at least include errors
so you can get a better idea what's going on.  Til then we're both
describing an elephant from opposite ends in the dark with blindfolds
on.