Thread: FATAL: could not read statistics message

We have been getting the following message in our log file since
upgrading to 8.0.3:

2005-07-20 21:24:15 BST [13299] FATAL:  could not read statistics
message: Resource temporarily unavailable

Does anyone know what this means?

The message seems to occur once every minute. We have the following
runtime statistics in our postgresql.conf

# - Statistics Monitoring -

#log_parser_stats = false
#log_planner_stats = false
#log_executor_stats = false
#log_statement_stats = false

# - Query/Index Statistics Collector -

#stats_start_collector = true
stats_command_string = true
#stats_block_level = false
#stats_row_level = false
#stats_reset_on_server_start = true

Also I notice that pg_stat_activity does not appear to refresh when I
query it to monitor current activity. I wonder if this is related?

Regards,

Alex Stanier.

Alexander Stanier <alexander.stanier@egsgroup.com> writes:
> 2005-07-20 21:24:15 BST [13299] FATAL:  could not read statistics
> message: Resource temporarily unavailable

> Does anyone know what this means?

It evidently means that the pgstat collector process got an EAGAIN errno
from recv() while trying to read a stats message.  While EAGAIN is
certainly documented as a possible failure from recv(), I'm finding it
hard to see why any kernel would return that for a read from a file that
has just told select() it is read-ready.  What platform are you on?
Were you having better luck with an earlier PG version, and if so which?

            regards, tom lane

We are on Mac OS X 10.3 (on an X-Serve G5) and we have just upgraded to
PostgreSQL 8.0.3 from 7.4.5. This problem was not evident in 7.4.5.

Regards,

Alex Stanier

Tom Lane wrote:

>Alexander Stanier <alexander.stanier@egsgroup.com> writes:
>
>
>>2005-07-20 21:24:15 BST [13299] FATAL:  could not read statistics
>>message: Resource temporarily unavailable
>>
>>
>
>
>
>>Does anyone know what this means?
>>
>>
>
>It evidently means that the pgstat collector process got an EAGAIN errno
>from recv() while trying to read a stats message.  While EAGAIN is
>certainly documented as a possible failure from recv(), I'm finding it
>hard to see why any kernel would return that for a read from a file that
>has just told select() it is read-ready.  What platform are you on?
>Were you having better luck with an earlier PG version, and if so which?
>
>            regards, tom lane
>
>
>