Thread: Restart time

Restart time

From
Jean Arnaud
Date:
Hello,

Is there a relation between database size and PostGreSQL restart duration ?
If true, i'm looking for a law predicting how much time is required to
restart PostGreSQL, depending on the DB size.

Does anyone now the behavior of restart time ?

Thanks

--
-- Jean Arnaud


Re: Restart time

From
Tom Lane
Date:
Jean Arnaud <Jean.Arnaud@inrialpes.fr> writes:
> Is there a relation between database size and PostGreSQL restart duration ?

No.

> Does anyone now the behavior of restart time ?

It depends on how many updates were applied since the last checkpoint
before the crash.

If you're talking about startup of a cleanly-shut-down database, it
should be pretty much constant time.

            regards, tom lane

Re: Restart time

From
"Rajesh Kumar Mallah"
Date:


On 12/5/06, Tom Lane <tgl@sss.pgh.pa.us> wrote:
Jean Arnaud <Jean.Arnaud@inrialpes.fr> writes:
> Is there a relation between database size and PostGreSQL restart duration ?

No.

> Does anyone now the behavior of restart time ?

It depends on how many updates were applied since the last checkpoint
before the crash.

If you're talking about startup of a cleanly-shut-down database, it
should be pretty much constant time.

Dear Sir,

Startup time of a clean shutdown database is constant. But we still
face problem when it comes to shutting down. PostgreSQL waits
for clients to finish gracefully. till date i have never been able to shutdown
quickly (web application scenerio) and i tend to do pg_ctl -m immediate stop
mostly.
 

                        regards, tom lane

---------------------------(end of broadcast)---------------------------
TIP 2: Don't 'kill -9' the postmaster

Re: Restart time

From
Tom Lane
Date:
"Rajesh Kumar Mallah" <mallah.rajesh@gmail.com> writes:
> Startup time of a clean shutdown database is constant. But we still
> face problem when it comes to shutting down. PostgreSQL waits
> for clients to finish gracefully. till date i have never been able to
> shutdown
> quickly (web application scenerio) and i tend to do pg_ctl -m immediate stop
> mostly.

RTFM ... you should be using -m fast not -m immediate.  -m immediate
is for emergency situations not routine.

            regards, tom lane

Re: Restart time

From
"Rajesh Kumar Mallah"
Date:


On 12/6/06, Tom Lane <tgl@sss.pgh.pa.us> wrote:
"Rajesh Kumar Mallah" <mallah.rajesh@gmail.com> writes:
> Startup time of a clean shutdown database is constant. But we still
> face problem when it comes to shutting down. PostgreSQL waits
> for clients to finish gracefully. till date i have never been able to
> shutdown
> quickly (web application scenerio) and i tend to do pg_ctl -m immediate stop
> mostly.

RTFM ... you should be using -m fast not -m immediate.  -m immediate
is for emergency situations not routine.

Thanks for correcting , -m fast  works fine for me.
I shall RTFM..... :)
Regds
mallah.                        regards, tom lane

Re: Restart time

From
Jean Arnaud
Date:
Rajesh Kumar Mallah a écrit :
>
>
> On 12/5/06, *Tom Lane* <tgl@sss.pgh.pa.us <mailto:tgl@sss.pgh.pa.us>>
> wrote:
>
>     Jean Arnaud <Jean.Arnaud@inrialpes.fr
>     <mailto:Jean.Arnaud@inrialpes.fr>> writes:
>     > Is there a relation between database size and PostGreSQL restart
>     duration ?
>
>     No.
>
>     > Does anyone now the behavior of restart time ?
>
>     It depends on how many updates were applied since the last checkpoint
>     before the crash.
>
>     If you're talking about startup of a cleanly-shut-down database, it
>     should be pretty much constant time.
>
>
> Dear Sir,
>
> Startup time of a clean shutdown database is constant. But we still
> face problem when it comes to shutting down. PostgreSQL waits
> for clients to finish gracefully. till date i have never been able to
> shutdown
> quickly (web application scenerio) and i tend to do pg_ctl -m
> immediate stop
> mostly.
>
>
>                             regards, tom lane
>
Hi

Thanks everybody for answers !

To be sure SGBD will stop before a certain time, I use function that is
a combination of pg_ctl -m fast to stop most of process cleanly, and
after few seconds, I send pg_ctl -m immediate to be shut down immediatly
the system if not already stoped. This works pretty well in practice and
offers a good compromise between clean and fast shutdown.

Cheers

--
--- Jean Arnaud
--- Projet SARDES
--- INRIA Rhône-Alpes