[Please keep the mailing list CC'd]
On Mon, Oct 11, 2010 at 8:40 PM, <Johannes.Meidert@rohde-schwarz.com> wrot=
e:
>
>> On Mon, Oct 11, 2010 at 2:23 AM, Craig Ringer
>> <craig@postnewspapers.com.au> wrote:
>> > On 10/08/2010 07:53 PM, Johannes Meidert wrote:
>> >>
>> >> The following bug has been logged online:
>> >>
>> >> Bug reference: =A0 =A0 =A05699
>> >> Logged by: =A0 =A0 =A0 =A0 =A0Johannes Meidert
>> >> Email address: =A0 =A0 =A0johannes.meidert@rohde-schwarz.com
>> >> PostgreSQL version: 9.0.1
>> >> Operating system: =A0 windows
>> >> Description: =A0 =A0 =A0 =A0Postgres runs only under default user.
>> >> Details:
>> >>
>> >> If a --superaccount is specified during installation other than
>> >> 'postgres',
>> >> the service will not start correctly and there are lots of errors "ro=
le
>> >> <postgres> =A0does not exist" in the DB error log.
>> >> The option -- superaccount should be removed.
>> >
>> > It does seem a bit odd - if the super account is wired into the databa=
se
>> > cluster at initdb time, you'd think it'd be read from there on startup.
>> >
>> > Comments, anybody? This report hasn't seen a response, but seems to ha=
ve
>> > some legitimacy as a usability issue if not an outright bug.
>>
>> I saw an email from our QA team a few minutes ago saying they have
>> been unable to reproduce the problem. What command line/options file
>> did you use exactly Johannes?
>
> The installation doesn't seem to be the problem. After installing we repl=
ace
> the
> pg_hba.conf with another one with 'trust' authentication:
>
> # IPv4 local connections:
> host =A0 =A0all =A0 =A0 =A0 =A0 all =A0 =A0 =A0 =A0 127.0.0.1/32 =A0 =A0 =
=A0 =A0 =A0trust
> # IPv4 all addresses. jm 2010
> host =A0 =A0all =A0 =A0 =A0 =A0 all =A0 =A0 =A0 =A0 0.0.0.0/0 =A0 =A0 =A0=
=A0 =A0trust
> # IPv6 local connections:
> #host =A0 =A0all =A0 =A0 =A0 =A0 all =A0 =A0 =A0 =A0 ::1/128 =A0 =A0 =A0 =
=A0 =A0 =A0 =A0 trust
>
> After that when starting the service there is an error message that the
> service
> didn't start 8though it is running) and we get the "role <postgres> does =
not
> exist"
> error messages.
> The error disappers if we change back to "MD5" or if we use user "postgre=
s"
> when installing.
>
> Probably for production systems it isn't a good idea anyway to use 'trust=
',
> so we will change.
>
> Installation:
>
> set PGDATA=3Dc:\_DB
>
> rem is using trust authentication, do not give a password. "trust" is set
> using the pg_hba.conf file
> postgresql-9.0.1-1-windows =A0--mode unattended --unattendedmodeui minimal
> --installer-language en --serviceaccount postgres --servicepassword
> system$01 --serverport 65432 --superaccount allaudio --servicename AudioDB
> --datadir %PGDATA%
>
> rem copy Config - files for remote - access
> copy *.conf %PGDATA%\ /y
>
> rem create DB.
> "%ProgramFiles%\PostgreSQL\9.0\bin\createdb" -e -p 65432 -h 127.0.0.1 -U
> allaudio AllAudio
>
>
I have a suspicion that the problem is that because your
serviceaccount is "postgres", pg_ctl tries to connect to the server as
"postgres" to test that it's up (the -w option). We've seen similar
problems in the past because the database name postgres is hardcoded
into pg_ctl.
[Get-out-of-jail card: having taken a suitable backup of your system],
can you edit the service command line in the registry and see if
startup then succeeds? You should find it in the key:
HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Services\AudioDB\ImagePath
--=20
Dave Page
Blog: http://pgsnake.blogspot.com
Twitter: @pgsnake
EnterpriseDB UK: http://www.enterprisedb.com
The Enterprise PostgreSQL Company