[7.4] statistics collector: Protocol not supported

On Sat, Nov 22, 2003 at 11:32:18PM -0400, Marc G. Fournier wrote:

I suspect it might be because I'm running in a jail'd environment, but
what should I be looking at to confirm?

could not create socket for statistics collector: Protocol not supported

You probably shouldn't worry about it. Maybe this message could
be made more clear too.

It probably does socket(AF_INET6), which fails, and gives the
above message, and then tries socket(AF_INET) which works.

Kurt

Kurt Roeckx <Q@ping.be> writes:

On Sat, Nov 22, 2003 at 11:32:18PM -0400, Marc G. Fournier wrote:

I suspect it might be because I'm running in a jail'd environment, but
what should I be looking at to confirm?

could not create socket for statistics collector: Protocol not supported

You probably shouldn't worry about it. Maybe this message could
be made more clear too.
It probably does socket(AF_INET6), which fails, and gives the
above message, and then tries socket(AF_INET) which works.

If you don't see something about "disabling statistics collector for
lack of working socket" then there's no reason to worry.

I'm surprised you don't see any similar message about failing to create
an IPv6 postmaster listen socket though?

regards, tom lane

On Sun, 23 Nov 2003, Tom Lane wrote:

Kurt Roeckx <Q@ping.be> writes:

On Sat, Nov 22, 2003 at 11:32:18PM -0400, Marc G. Fournier wrote:

I suspect it might be because I'm running in a jail'd environment, but
what should I be looking at to confirm?

could not create socket for statistics collector: Protocol not supported

You probably shouldn't worry about it. Maybe this message could
be made more clear too.
It probably does socket(AF_INET6), which fails, and gives the
above message, and then tries socket(AF_INET) which works.

If you don't see something about "disabling statistics collector for
lack of working socket" then there's no reason to worry.

I'm surprised you don't see any similar message about failing to create
an IPv6 postmaster listen socket though?

Nope, this is everything from the startup:

Nov 23 03:30:58 pgsql74 postgres[59990]: [1-1] LOG: could not create socket for statistics collector: Protocol not supported
Nov 23 03:30:58 pgsql74 postgres[59993]: [2-1] LOG: database system was shut down at 2003-11-23 03:22:39 GMT
Nov 23 03:30:58 pgsql74 postgres[59993]: [3-1] LOG: checkpoint record is at 0/9B0B8C
Nov 23 03:30:58 pgsql74 postgres[59993]: [4-1] LOG: redo record is at 0/9B0B8C; undo record is at 0/0; shutdown TRUE
Nov 23 03:30:58 pgsql74 postgres[59993]: [5-1] LOG: next transaction ID: 534; next OID: 17142
Nov 23 03:30:58 pgsql74 postgres[59993]: [6-1] LOG: database system is ready

Is it easy to do a quick message change to state that it was the IPv6
socket that failed?

----
Marc G. Fournier Hub.Org Networking Services (http://www.hub.org)
Email: scrappy@hub.org Yahoo!: yscrappy ICQ: 7615664

"Marc G. Fournier" <scrappy@postgresql.org> writes:

Is it easy to do a quick message change to state that it was the IPv6
socket that failed?

I deliberately did not do that because we were well past error message
freeze for 7.4, but we certainly should fix it for 7.5.

regards, tom lane

Tom Lane wrote:

"Marc G. Fournier" <scrappy@postgresql.org> writes:

Is it easy to do a quick message change to state that it was the IPv6
socket that failed?

I deliberately did not do that because we were well past error message
freeze for 7.4, but we certainly should fix it for 7.5.

Yes, I am unsure why the mention of IPv6 was removed from the original
patch I submitted. Does anyone remember?

-- 
  Bruce Momjian                        |  http://candle.pha.pa.us
  pgman@candle.pha.pa.us               |  (610) 359-1001
  +  If your life is a hard drive,     |  13 Roberts Road
  +  Christ can be your backup.        |  Newtown Square, Pennsylvania 19073

Bruce Momjian <pgman@candle.pha.pa.us> writes:

Yes, I am unsure why the mention of IPv6 was removed from the original
patch I submitted. Does anyone remember?

You sure you aren't confusing this code with the listen-socket-opening
code over in backend/libpq/ ?

The libpq code identifies the socket type (and I think we should just
borrow that logic for pgstat.c) but I don't believe anyone proposed
making pgstat do so, because up till this recent patch it wouldn't try
more than one socket type anyway.

regards, tom lane

Tom Lane wrote:

Bruce Momjian <pgman@candle.pha.pa.us> writes:

Yes, I am unsure why the mention of IPv6 was removed from the original
patch I submitted. Does anyone remember?

You sure you aren't confusing this code with the listen-socket-opening
code over in backend/libpq/ ?

The libpq code identifies the socket type (and I think we should just
borrow that logic for pgstat.c) but I don't believe anyone proposed
making pgstat do so, because up till this recent patch it wouldn't try
more than one socket type anyway.

Oh, yes confused. I never considered the stats listener message before.

-- 
  Bruce Momjian                        |  http://candle.pha.pa.us
  pgman@candle.pha.pa.us               |  (610) 359-1001
  +  If your life is a hard drive,     |  13 Roberts Road
  +  Christ can be your backup.        |  Newtown Square, Pennsylvania 19073