Uhhhhn, sad.
There is one last suggestion I can make.
I have written a SELECT query in protocols/pgsql.c and it is
comment-ou
t'ed. I made sure it is left in Monit-4.9 source code. Then, uncomment
the section starting with
/** Doing this is too suspicious maybe.
and under the line that says
/** This is where suspicious people can do SELECT query that I
dont
*/
insert
socket_write(s, (unsigned char *)requestQuery, sizeof(requestQuer
y));
Then do make.
This will make monit do a REAL JOB so PostgreSQL may satisfy
itself. To
see these lines are processed, edit postgresql.conf to include
these s
ettings and restart PostgreSQL;
log_connections = yes
log_statement = yes
(on PostgreSQL 8.x, set log_statement to `all')
If the queries `select 1 as a' dont appear in postgresql log,
something
totally wrong is going on and I have to give up.
I hope this helps solve the problem.
Tatsuya Nonogaki
~//^^//^^//^^//^^//^^//^^//^^//^^//^^//^^//^^//^^//^^//^^//^^//
Hi Nonogaki san,
Thank you for your prompt response.
I've tried different combinations as you suggest. I even tried
the unixso
cket check as you suggested, but am still getting these in my
pgsql log fil
e:
2007-09-07 01:05:50 LOG: incomplete startup packet
2007-09-07 01:10:50 LOG: incomplete startup packet
I am using Postgresql version 7.4.7
I think the test is working properly, as it is able to connect to
the dat
abase and make sure the connection is possible.
On searching a bit on the web (e.g., see <http://www.mail-
archive.com/pgs
address@hidden/msg20274.html>) I found out that this
could be happ
ening because the monitoring client is disconnecting too quickly
for postgr
esql's liking.
Raj
On Sep 6, 2007, at 8:56 PM, Tatsuya Nonogaki wrote:
Hi,Raj.
I suspect monit is actually using pgsql test. Bellow are would-be
list
of possibility.
* Are TEST type names case insensitive and `port' optional? Try
this (I
am not sure...)
if failed port 5432 protocol pgsql
then ...
* PGSQL TEST was written for front-end protocol version 3.0 which
had
been implemented since PostgreSQL 7.4. If you are using pre-7.4
PostgreSQL server, the TEST may not work properly.
* Does your PostgreSQL allow root to connect to `root' DB you have
created? Ensure pg_hba.conf has something like;
host root root 127.0.0.1/32 trust
local root root ident sameuser
(Note that this pg_hba.conf format is based on PostgreSQL 8.x. It
varies
on 7.4.x)
The second line is needed for pgsql TEST through UNIX Socket and
unnecessarry for TCP. But I personally recommend using UNIX Socket
unless you REALLY need to examine PostgreSQL availability through
TCP.
In that case, monit definition should be;
if failed unixsocket /tmp/.s.PGSQL.5432 protocol pgsql
then ...
Best regards.
Tatsuya Nonogaki
Hi,
I've just downloaded monit and have been playing around with it
to star
t
monitoring our Apache/Tomcat/Postgresql based web application.
(Since the application went down for a few hours this weekend,
we've be
en
looking for an automated re-start solution, and are glad to have
found
monit.)
When checking for Postgresql I've been using the PGSQL entry
like this:
check process postgresql with pidfile /var/run/postmaster.5432.pid
start program = "/etc/init.d/postgresql start"
stop program = "/etc/init.d/postgresql stop"
if failed protocol PGSQL then restart
if 5 restarts within 5 cycles then timeout
I've also added the user root and database root to avoid having
spuriou
s
messages in my pglog file complaining about user root and
database root.
I still keep getting errors in the PGLOG file like:
LOG: incomplete startup packet
Can someone suggest a way to not get these errors. Since I am
monitorin
g
every 2 minutes this will create too many messages in our log
file and I
am
worried we might miss some other critical messages.
Raj
--
Raj Gupta address@hidden
1684 Nightingale Avenue Suite 201
Sunnyvale, CA 94087 408-733-2737(fax)
http://www.zeemaps.com
--
To unsubscribe:
http://lists.nongnu.org/mailman/listinfo/monit-general
--
To unsubscribe:
http://lists.nongnu.org/mailman/listinfo/monit-general
--
To unsubscribe:
http://lists.nongnu.org/mailman/listinfo/monit-general
--
To unsubscribe:
http://lists.nongnu.org/mailman/listinfo/monit-general