Обсуждение: hot restart of posgtresql

hi all
by unknown reason our database server rebooted couple hours ago.
now i cannot restart posgresql 8.1
heres the error code i get:


=====================================================
root@eddie:/var/run# /etc/init.d/postgresql-8.1 start
Starting PostgreSQL 8.1 database server: mainThe PostgreSQL server failed
to start. Please check the log output:
2007-07-03 19:20:42 EDT  2928 FATAL:  could not create shared memory
segment: Invalid argument
2007-07-03 19:20:42 EDT  2928 DETAIL:  Failed system call was
shmget(key=5432001, size=3276259328, 03600).
2007-07-03 19:20:42 EDT  2928 HINT:  This error usually means that
PostgreSQL's request for a shared memory segment exceeded your kernel's
SHMMAX parameter.  You can either reduce the request size or reconfigure
the kernel with larger SHMMAX.  To reduce the request size (currently
3276259328 bytes), reduce PostgreSQL's shared_buffers parameter (currently
393216) and/or its max_connections parameter (currently 250).
        If the request size is already small, it's possible that it is
less than your kernel's SHMMIN parameter, in which case raising
the request size or reconfiguring SHMMIN is called for.
        The PostgreSQL documentation contains more information about
shared memory configuration.
 failed!
=======================================================

heres my sysctl:
kernel.shmmax=4294967296
kernel.shmall=536870912


Any help will be really appreciated
Thanks

ngaleyev@wavelengthmail.com writes:
> 2007-07-03 19:20:42 EDT  2928 FATAL:  could not create shared memory
> segment: Invalid argument
> 2007-07-03 19:20:42 EDT  2928 DETAIL:  Failed system call was
> shmget(key=5432001, size=3276259328, 03600).
> 2007-07-03 19:20:42 EDT  2928 HINT:  This error usually means that
> PostgreSQL's request for a shared memory segment exceeded your kernel's
> SHMMAX parameter.  You can either reduce the request size or reconfigure
> the kernel with larger SHMMAX.  To reduce the request size (currently
> 3276259328 bytes), reduce PostgreSQL's shared_buffers parameter (currently
> 393216) and/or its max_connections parameter (currently 250).

> heres my sysctl:
> kernel.shmmax=4294967296
> kernel.shmall=536870912

Why have you got SHMALL set to half a GB when you have Postgres
configured to ask for over 3 GB?

            regards, tom lane

Tom Lane wrote:
> ngaleyev@wavelengthmail.com writes:
> > 2007-07-03 19:20:42 EDT  2928 FATAL:  could not create shared memory
> > segment: Invalid argument
> > 2007-07-03 19:20:42 EDT  2928 DETAIL:  Failed system call was
> > shmget(key=5432001, size=3276259328, 03600).
> > 2007-07-03 19:20:42 EDT  2928 HINT:  This error usually means that
> > PostgreSQL's request for a shared memory segment exceeded your kernel's
> > SHMMAX parameter.  You can either reduce the request size or reconfigure
> > the kernel with larger SHMMAX.  To reduce the request size (currently
> > 3276259328 bytes), reduce PostgreSQL's shared_buffers parameter (currently
> > 393216) and/or its max_connections parameter (currently 250).
>
> > heres my sysctl:
> > kernel.shmmax=4294967296
> > kernel.shmall=536870912
>
> Why have you got SHMALL set to half a GB when you have Postgres
> configured to ask for over 3 GB?

Hmm, isn't shmall measured in pages?

I think the problem may be that he changed the settings in sysctl.conf
but he hasn't rebooted since, so the kernel is running with other
settings.  Trying with sysctl -w to change the actual values might prove
helpful.

--
Alvaro Herrera                                http://www.CommandPrompt.com/
The PostgreSQL Company - Command Prompt, Inc.

Alvaro Herrera <alvherre@commandprompt.com> writes:
> Tom Lane wrote:
>> ngaleyev@wavelengthmail.com writes:
>>> heres my sysctl:
>>> kernel.shmmax=4294967296
>>> kernel.shmall=536870912
>>
>> Why have you got SHMALL set to half a GB when you have Postgres
>> configured to ask for over 3 GB?

> Hmm, isn't shmall measured in pages?

Depends on platform, which he didn't say ... but if it is a platform
that measures in pages, I wonder whether that value is provoking an
integer overflow somewhere.  I don't think Linux allows more than
about 2 million, for instance.

> I think the problem may be that he changed the settings in sysctl.conf
> but he hasn't rebooted since, so the kernel is running with other
> settings.  Trying with sysctl -w to change the actual values might prove
> helpful.

But his DB *was* running before, so somehow the active sysctl settings
are different now than they were.  I was betting on there having been
a manual tweak to sysctl that never got made in the configuration file,
and thus got lost on reboot.

            regards, tom lane

thanks for replies everyone
platform is Debian 2.6 btw

Heres how i got temp fix (i dont have much background in postgresql, so dont
kick me hard):
few months ago old admin upgraded to 6GB of ram and increased shared_buffers
in posgresql.conf from  16384 to 393216 (3GB)
He also modifed sysctl to look like this:
-----
kernel.shmmax=4294967296
kernel.shmall=536870912
-----
I looked in top command, and dmesg, both indicate 6GB of ram. So I started
gradually increase "sysctl -w kernel.shmmax=xxx"
during that i was checking out  /proc/sys/kernel/shmmax
it was keeping up with corresponding values until i reached that number -
4294967296, it was showing "0" instead. so i lowered it to 4200000000,
about highest number that would not null that /proc value.
Still wouldnt start :( but with different error now, didnt save it
After that i decreased shared_buffers back to 16384, and it worked. Didnt
mess with it anymore for right now, don't want to experiment on live server.
Old admin is not around anymore, so I sort of inherited responsibility for
the server, but I don't know much about postgresql yet.

So, whats the optimum values for this amount of memory as far as SHMALL, MAX
and shared_buffers?
Database size is about 115GB, daily traffic is about 800MB.

Thanks alot
-Nik