Thread: Error: "could not read from statistics collector pipe"

Hello!

In my PostgreSQL logfile, I see that the following entries occur
very frequently:

2006-01-27 10:37:29 FATAL  could not read from statistics collector
                           pipe: No error
2006-01-27 10:37:30 LOG    statistics collector process (PID 5940)
                           was terminated by signal 1

(PostgreSQL 8.1.2, WinXP SP2, Athlon64 X2 3800, 2 GB RAM)

These errors occur every 1-2 minutes when the DB is active. What do
these messages mean and how can I stop them from appearing?

I also noticed that during inserting lots of records, the DB becomes
increasingly slower the more records were inserted. For example, the
first 100000 records take 15 minutes, but records 300000-400000 take
3 hours. Could this be related to the messages above?

I tried to execute ANALYZE every 100000 records, but this did not
seem to help.

Thanks in advance for your help!

- Stephan

> 2006-01-27 10:37:29 FATAL  could not read from statistics collector
>                            pipe: No error
> 2006-01-27 10:37:30 LOG    statistics collector process (PID 5940)
>                            was terminated by signal 1
>

try turning row level statistics off (or, during bulk inserts, turn
inserts off completely).  I've seem this too but never nailed it down.

> (PostgreSQL 8.1.2, WinXP SP2, Athlon64 X2 3800, 2 GB RAM)
>
> These errors occur every 1-2 minutes when the DB is active. What do
> these messages mean and how can I stop them from appearing?
>
> I also noticed that during inserting lots of records, the DB becomes
> increasingly slower the more records were inserted. For example, the
> first 100000 records take 15 minutes, but records 300000-400000 take
> 3 hours. Could this be related to the messages above?

try defering index/key generationg until after the insert.  probably
what is happening is you are blowing out your sort memory and swapping
occurs.  of course, you can always bump sort_mem

merlin

Sorry for the late reply, I didn't have much time during the last
days to catch up on this list.

Merlin Moncure wrote:

> try turning row level statistics off (or, during bulk inserts, turn
> inserts off completely).  I've seem this too but never nailed it down.

Thanks, I'll give that a try!

- Stephan