Have 14 systems scattered across the country where... hardware is
identical, all run linux OS RHE4.0 with postgresql v7.4.8.
The hardware is about 4 years old and is scheduled for replacement
sometime next year.
The database structure is identical on each system, the difference is in
the data that is stored within each database, as it varies from site to
site and the amount of data varies as well, so no two databases are
exactly the same size or contains the same data.
Every system runs a "vacuum analyze" 2 to 3 times per day on this
database and a nightly cron job does a pg_dump of the database.

One site is getting errors with the "vacuum analyze" so it basically
fails and similar error then occurs when the nightly backups is run.
The error in the vacuum's logfile looks like this:
INFO: vacuuming "public.pecrsep"

vacuumdb: vacuuming of database "adb_ob72orn" failed: ERROR: could not
read block 18658 of relation "pecrsep_time_ind": Input/output error

The Elec. Technicians at first thought it was an issue with one of the
hard drives, but a very thorough check of the hardware has eliminated
this possibility.
So far they have tried re-installing everything as if its the first time
from scratch on the existing hardware with no success. As a last
resort, the hardware was replaced with a spare (similar age) and
everything installed from scratch. The database and related apps ran
fine for about a week and then the same type of error occurred again
with vacuumdb and pg_dump and it appears to be occurring with the same
database table.

At this point, we don't know what else to try, and the problem hasn't
occurred anywhere else (fortunately). Does anyone have any idea what
this error means and how to fix it.

Any suggestions, ideas, etc would be greatly appreciated.
Julie Meyer




---------------------------(end of broadcast)---------------------------
TIP 2: Don't 'kill -9' the postmaster

Juliann Meyer <Julie.Meyer@noaa.gov> writes:
> vacuumdb: vacuuming of database "adb_ob72orn" failed: ERROR: could not
> read block 18658 of relation "pecrsep_time_ind": Input/output error

This says read() failed with errno EIO, ie the operating system reported
a hardware failure while trying to read that disk block.

> The Elec. Technicians at first thought it was an issue with one of the
> hard drives, but a very thorough check of the hardware has eliminated
> this possibility.

Not thorough enough I would say, because there is no other cause of that
error code than hardware problems.

You do know that PG 7.4.8 is nine minor releases behind the times?
That's not the cause of your immediate problem, but personally
I'd suggest an update before some other bug bites you. See
http://developer.postgresql.org/pgdo...s/release.html

regards, tom lane

---------------------------(end of broadcast)---------------------------
TIP 2: Don't 'kill -9' the postmaster