On Thu, Jan 06, 2011 at 12:18:12AM +0100, Janez Barbič wrote:
Hi Matthew,
I solved it! :)
Wonderful.
Even though I noticed zeroed out port I paid no attention to it because I assumed syslog-ng used default Postgres port. Postgres is listening to port 5432 (default), but syslog-ng was sending packets to port 0
Normally something like this would not happen, but there have been a series of odd bugs in the behavior of DB ports for different DBs due to various syslog-ng vs. DBI interactions that have gone wrong.
(again, thanks for Wireshark idea).
I used to create network anomaly detection software, so I never believe anything about socket programming unless I have packet captures or detailed debug logs. Preferably both, because firewalls and applications can reject traffic at L3-L7 after the packet capture gets the traffic at L2.
So I just pointed syslog-ng to the correct port and it started to work.
Good thing it did. Otherwise we would have had to track down another port bug. Although it's a bug it defaults to port 0 which nobody uses instead of defaulting to the Postgres port. Maybe you could put this into Bugzilla?
I must also say that I am positively surprised by really fast response :)
There's a rule of open source that if you want a prompt response on a mailing list, you should provide prompt responses to everyone else. I really try to follow this.
Best regards, Janez Barbic
Regards, Matthew Hall.