"Incomplete Startup Packet" debug log message on startup

[Jennifer Redman]

Binary package hint: postgresql-8.3

Running Ubuntu 8.04 LTS, with postgresql-8-3, postgresql-client-8.3, postgresql-common - standard Ubuntu stable postgresql package. The same error occurs on a system with postgresql-8-2. Both boxes are also running python-psycopg2. Connections limited to localhost. Nothing monitoring or scanning port 5432 at a 5 hour interval.

I've checked 2 other Ubuntu systems running the standard postgresql package -- (Fiesty with postgresql 8.2 and another Hardy system with 8.3) all have the same error messages.

The message appears on initial startup after package installation before any custom configuration changes are made, and continues after custom configs are made.

I'm get the following error in the logs at startup and at the exact interval that -x /usr/sbin/pg_maintenance ]; then /usr/sbin/pg_maintenance --analyze appears to run in cron:

From /postgresql-8.3-main.log :

2008-12-17 00:02:01 PST LOG: incomplete startup packet
2008-12-17 05:02:01 PST LOG: incomplete startup packet
2008-12-17 10:02:01 PST LOG: incomplete startup packet
2008-12-17 15:02:01 PST LOG: incomplete startup packet
2008-12-17 20:02:01 PST LOG: incomplete startup packet

Verbose on startup:

2008-12-18 12:15:48 PST LOCATION: StartupXLOG, xlog.c:4816
2008-12-18 12:15:48 PST LOG: 00000: autovacuum launcher started
2008-12-18 12:15:48 PST LOCATION: AutoVacLauncherMain, autovacuum.c:520
2008-12-18 12:15:48 PST LOG: 00000: database system is ready to accept connections
2008-12-18 12:15:48 PST LOCATION: reaper, postmaster.c:2156
2008-12-18 12:15:49 PST LOG: 08P01: incomplete startup packet
2008-12-18 12:15:49 PST LOCATION: ProcessStartupPacket, postmaster.c:1396

From /syslog

Dec 17 00:02:01 nameofserver /USR/SBIN/CRON[26124]: (root) CMD (if [ -x /usr/sbin/pg_maintenance ]; then /usr/sbin/pg_maintenance --analyze >/dev/null; fi)
Dec 17 05:02:01 nameofserver /USR/SBIN/CRON[1879]: (root) CMD (if [ -x /usr/sbin/pg_maintenance ]; then /usr/sbin/pg_maintenance --analyze >/dev/null; fi)
Dec 17 10:02:01 nameofserver /USR/SBIN/CRON[9264]: (root) CMD (if [ -x /usr/sbin/pg_maintenance ]; then /usr/sbin/pg_maintenance --analyze >/dev/null; fi)
Dec 17 15:02:01 nameofserver /USR/SBIN/CRON[15886]: (root) CMD (if [ -x /usr/sbin/pg_maintenance ]; then /usr/sbin/pg_maintenance --analyze >/dev/null; fi)
Dec 17 20:02:01 nameofserver /USR/SBIN/CRON[12183]: (root) CMD (if [ -x /usr/sbin/pg_maintenance ]; then /usr/sbin/pg_maintenance --analyze >/dev/null; fi)

[Jennifer Redman (jenred)]

UPDATE:

Running
# /etc/init.d/postgresql-8.3 autovac-restart
 * Restarting PostgreSQL 8.3 autovacuum daemon

Results in:
 * Error: PostgreSQL 8.1 and above has an integrated autovacuum daemon which cannot be controlled by this program.

Looks like /etc/cron.d/postgresql-common

# Run VACUUM ANALYSE on all databases every 5 hours if pg_autovacuum is not
# running
2 0,5,10,15,20 * * 1-6 root if [ -x /usr/sbin/pg_maintenance ]; then /usr/sbin/pg_maintenance --analyze >/dev/null; fi

Is deprecated for postgresql versions 8.1 and higher and needs to be commented out by default in those packages.

The same error message appears on start-up
( incomplete startup packet)

but I could not figure out where in /usr/share/postgresql-common/init.d-functions the pg_maintenance script was being called.

Additionally,

Running:

#/sbin/pg_maintenance --analyze

results in:
Skipping cluster 8.3/main since autovacuuming is enabled for it

and the subsequent:
2008-12-18 15:21:18 PST LOG: 08P01: incomplete startup packet
2008-12-18 15:21:18 PST LOCATION: ProcessStartupPacket, postmaster.c:1396

in /postgresql-8.3-main.log :

I'd submit a patch if I could figure out where this is being called on init -- but alas, can't figure this part out. But it does appear that the pg_maintenance script needs to be disabled for Postgresql 8.1 and higher or a version check needs to be added.

[Christian Ehrhardt  (paelzer)]

The way this resolved in Debian applies to us as well.
This message isn't gone but much more uncommon these days - so it is less of an issue or rather less severe.

Christoph still wants to send a patch upstream to make it a real debug message that does not show up at all by default, but for the initial issue of "logs being full of these messages" that can be considered fixed already.