Am 08.08.2014 um 01:20 schrieb Tom Lane:
> Andreas <maps.on@gmx.net> writes:
>> When I checked the logfile of the local PG-9.3 server I found an error
>> on startup that complained it couldn't bind a IPv4 socket for 5432.
>> Another error says autovauum is not running.
>
> It would help if you provided the exact log messages, but what this sounds
> like is a networking configuration problem. Do you have DNS service on
> this machine, and if so does it correctly resolve "localhost"?
>
> There has been some talk in the past about kluging things so that PG would
> work even if "localhost" didn't resolve, but the case hasn't seemed worth
> putting much effort into.
Hi Tom,
I guess I partly figured it out.
1) The big issue was that when I sat up eth0 to a fixed IP, I
accidentally commented out the config line for LO. Therefore there was
no known protocol for the socket and PG couldn't bind it to localhost.
2) It appeares that when LO is set to INET (IPv4) it doesn't work as
IPv6. On the other hand PG with the provided postgresql.conf tries to
listen on "localhost" and interprets this as an IPv6 address.
This prohibits the 1'st generation of the cluster while initial
installation through aptitude.
After changing the listen_adresses to "127.0.0.1" I could generate the
cluster and PG works.
3) There still is some minor glitch in the startup of the PG service.
Even with a complete new install of Raspian + dist-upgrade and
installation of postgres-9.1 from the repository, I get a line like:
"CEST LOG: incomplete startup packet"
in the log, everytime I start up PG. It doesn't repeat later.
This doesn't go away with the self compiled PG 9.3.5 so it may rather be
triggered by something in Raspian.
Otherwise PG seems to run now ... well rather "move" as it's on a 700MHz
singlecore ARM :}
Do you have a suggestion about #3 ?
Regards
Andreas