Обсуждение: [HACKERS] Bringing down postmaster at regular intervals.

Поиск
Список
Период
Сортировка

[HACKERS] Bringing down postmaster at regular intervals.

От
David Friend
Дата:
On Fri, 6 Jun 1997, Marc G. Fournier wrote:

> On Fri, 6 Jun 1997, David Friend wrote:

> > I am thinking of adding a comment in the INSTALL instructions suggesting
> > that the back end should be stopped and restarted on a regular basis as
> > part of the standard maintainence procedures.  I will also mention that
> > this will probably not be necessary once our next release comes out.
>
>     Hrmmm...what is 'regular basis'?  Once a week?  A day?  A month?
> 'regular basis', I think, is much too vague :(

Given that Tom and I both started noticing serious problems after about
4-6 runs of the regression tests, I was thinking of suggesting that an
adiministrator guestimate how long it would take his system to generate a
similar amount of traffic.  Nightly or weekly would be my suggestion.
Also suggest that it should be done whenever the system starts swapping a
lot.

The simple thing is to write a shell script that is run nightly or weekly
by cron.  The shell script could bring down the postmaster, alter the
permissions so that only the postgress account can use it, bring up
postmaster, do a backup (include something to keep the last few backups
around), then bring down the postmaster, restore the permissions to
normal, and bring it up again.

Any volunteers?  <grin>

David Friend                ! cq995@freenet.carleton.ca
Atlantis Scientific Inc.        ! david.friend@atlsci.com
20 Colonnade Rd, Suite 110        ! 613-727-1087 (voice)
Ottawa, Ontario, CANADA  K2E 7M6    ! 800-265-3894 (voice)
ERGOvista Scientific Image Analysis    ! 613-727-5853 (fax)

------------------------------

Re: [HACKERS] Bringing down postmaster at regular intervals.

От
Bruce Momjian
Дата:
>
> On Fri, 6 Jun 1997, Marc G. Fournier wrote:
>
> > On Fri, 6 Jun 1997, David Friend wrote:
>
> > > I am thinking of adding a comment in the INSTALL instructions suggesting
> > > that the back end should be stopped and restarted on a regular basis as
> > > part of the standard maintainence procedures.  I will also mention that
> > > this will probably not be necessary once our next release comes out.
> >
> >     Hrmmm...what is 'regular basis'?  Once a week?  A day?  A month?
> > 'regular basis', I think, is much too vague :(
>
> Given that Tom and I both started noticing serious problems after about
> 4-6 runs of the regression tests, I was thinking of suggesting that an
> adiministrator guestimate how long it would take his system to generate a
> similar amount of traffic.  Nightly or weekly would be my suggestion.
> Also suggest that it should be done whenever the system starts swapping a
> lot.
>
> The simple thing is to write a shell script that is run nightly or weekly
> by cron.  The shell script could bring down the postmaster, alter the
> permissions so that only the postgress account can use it, bring up
> postmaster, do a backup (include something to keep the last few backups
> around), then bring down the postmaster, restore the permissions to
> normal, and bring it up again.

from what Vadim said, I am inclined to believe that the problem with the
regression tests is that it uses some very obscure commands to test that
would not normally be used in normal use, so the leaks from the
regression tests may not be an accurate indication that restarts are
necessary.

Marc is running 6.1 beta on his live system.  Is anyone seeing problems
with running 6.1 for a long time under typical loads?

- --
Bruce Momjian
maillist@candle.pha.pa.us

------------------------------

Re: [HACKERS] Bringing down postmaster at regular intervals.

От
Igor
Дата:
I've been running 1.09 beta  :( on a live system for the past year,
restarting the postmaster only when the machine crashed ... there were no
memory problems that required either postmaster restart or anything
else...of course it was running on Digital Unix (whose memory management
is very good). I am running the same thing now with 6.1 beta, and no
problems anyway... Of course if you take a workstation, or some machine
with very little ram, it might be a problem....

=+=------------------------/\---------------------------------=+=
       Igor Natanzon      |**|   E-mail: igor@sba.miami.edu
=+=------------------------\/---------------------------------=+=

On Fri, 6 Jun 1997, Bruce Momjian wrote:

> >
> > On Fri, 6 Jun 1997, Marc G. Fournier wrote:
> >
> > > On Fri, 6 Jun 1997, David Friend wrote:
> >
> > > > I am thinking of adding a comment in the INSTALL instructions suggesting
> > > > that the back end should be stopped and restarted on a regular basis as
> > > > part of the standard maintainence procedures.  I will also mention that
> > > > this will probably not be necessary once our next release comes out.
> > >
> > >     Hrmmm...what is 'regular basis'?  Once a week?  A day?  A month?
> > > 'regular basis', I think, is much too vague :(
> >
> > Given that Tom and I both started noticing serious problems after about
> > 4-6 runs of the regression tests, I was thinking of suggesting that an
> > adiministrator guestimate how long it would take his system to generate a
> > similar amount of traffic.  Nightly or weekly would be my suggestion.
> > Also suggest that it should be done whenever the system starts swapping a
> > lot.
> >
> > The simple thing is to write a shell script that is run nightly or weekly
> > by cron.  The shell script could bring down the postmaster, alter the
> > permissions so that only the postgress account can use it, bring up
> > postmaster, do a backup (include something to keep the last few backups
> > around), then bring down the postmaster, restore the permissions to
> > normal, and bring it up again.
>
> >From what Vadim said, I am inclined to believe that the problem with the
> regression tests is that it uses some very obscure commands to test that
> would not normally be used in normal use, so the leaks from the
> regression tests may not be an accurate indication that restarts are
> necessary.
>
> Marc is running 6.1 beta on his live system.  Is anyone seeing problems
> with running 6.1 for a long time under typical loads?
>
> --
> Bruce Momjian
> maillist@candle.pha.pa.us
>

------------------------------