How serious are these problems?

Started by Heatsink Adminalmost 20 years ago3 messagesgeneral
Jump to latest
#1Heatsink Admin
admin@heatsinksoftware.com

Hi,

We have been having some problems with a postgres database recently and would appreciate any advice or pointers.

Environment:
Postgres 8.1.4 running on Windows Server 2003 Enterprise x64 Edition (installed using standard win32 installer).
Quad (2 x dual core) AMD Opteron 870. 8Gb RAM. 8x250Gb SATA Raid 10 array.
Main database is around 30Gb � dominated by 2 large tables (one 7m rows using 11Gb, and another 55m rows using 9gb). Encoding is UTF8.

Symptoms:
Two separate symptoms � may be completely unconnected?

1) When running backups using pg_dump (from pgAdmin � ie using �f option for output), about half way through the �dumping contents of table x� stage I get the message :

pg_dump: [custom archiver] WARNING: ftell mismatch with expected position -- ftell used

repeated for every table afterwards. When trying to restore this backup I get :

pg_restore: [custom archiver] error during file seek: Invalid argument

and running the pg_dump with the TAR option I get :

pg_dump: [tar archiver] actual file length (591516459) does not match expected (591516459)

I did manage to run an error/warning free pg_dump using the plain option � but I haven�t yet proved this restores cleanly (can someone confirm the proper way to do this since pgAdmin doesn�t seem to support it � I guess command line psql, but flags etc?).

I�m hoping that this problem is isolated to usage of pg_dump/pg_restore in non-plain mode (maybe affected by Win 2003 x64 ?) and not symptomatic of anything fundamental � however the next problem makes me worry about that�

2) Occasionally (3 times in the last 2 months) the database locks up. When this happens the log file contains a bunch of errors along the lines of :

2006-06-26 14:53:40 LOG: could not fsync segment 0 of relation 1663/16672/50807554: Permission denied

2006-06-26 14:53:40 ERROR: storage sync failed on magnetic disk: Permission denied

Bouncing the postgres service seems to get everything working again? I�ve checked that the postgres account has permissions on all the files, and the last couple of times this happened I ran a forced re-index on everything (following a suggestion on a similar problem in another thread).

Any advice on these issues would be greatly appreciated.

Thanks � Barry

#2Martijn van Oosterhout
kleptog@svana.org
In reply to: Heatsink Admin (#1)
Re: How serious are these problems?

On Mon, Jul 03, 2006 at 06:57:03AM -0700, Steven Ridgway wrote:

Hi,
We have been having some problems with a postgres database recently and would appreciate any advice or pointers.

Environment:
Postgres 8.1.4 running on Windows Server 2003 Enterprise x64 Edition (installed using standard win32 installer).
Quad (2 x dual core) AMD Opteron 870. 8Gb RAM. 8x250Gb SATA Raid 10 array.
Main database is around 30Gb – dominated by 2 large tables (one 7m rows using 11Gb, and another 55m rows using 9gb). Encoding is UTF8.

Symptoms:
Two separate symptoms – may be completely unconnected?

1) When running backups using pg_dump (from pgAdmin – ie using –f option for output), about half way through the “dumping contents of table x” stage I get the message :

pg_dump: [custom archiver] WARNING: ftell mismatch with expected position -- ftell used

repeated for every table afterwards. When trying to restore this backup I get :

pg_restore: [custom archiver] error during file seek: Invalid argument

and running the pg_dump with the TAR option I get :

pg_dump: [tar archiver] actual file length (591516459) does not match expected (591516459)

This appears to be some kind of 64-bit issue. ftell is probably not
returning a 64-bit offset. The print is obviously truncating the length
also.

Not sure what can be done about this, any Windows hackers about?

Have a nice day,
--
Martijn van Oosterhout <kleptog@svana.org> http://svana.org/kleptog/

Show quoted text

From each according to his ability. To each according to his ability to litigate.

#3Tom Lane
tgl@sss.pgh.pa.us
In reply to: Heatsink Admin (#1)
Re: How serious are these problems?

Steven Ridgway <admin@heatsinksoftware.com> writes:

2) Occasionally (3 times in the last 2 months) the database locks up. When this happens the log file contains a bunch of errors along the lines of :

2006-06-26 14:53:40 LOG: could not fsync segment 0 of relation 1663/16672/50807554: Permission denied

2006-06-26 14:53:40 ERROR: storage sync failed on magnetic disk: Permission denied

We've seen reports of this sort of symptom due to anti-virus software
deciding that the postmaster is a virus :-(. What add-on AV code do
you have on that machine?

regards, tom lane