Обсуждение: PostgreSQL 8.0.1 problem
Hi: I want migrate to PostgreSQL 8.0.1 for Windows from PostgreSQL 7.4.1, but I have a big problem. In 7.4.1 I have a large view with nested views. It lasts 1 minute and 20 seconds to finish. In 8.0.1 for Windows it lasts.... I don't know. One hour, two hour, one day... I always cancel the view because it lasts forever. Shared buffers and work mem in 8.0.1 are greater than those of 7.4.1. So I think this isn't a memory problem. Furthermore, the 8.0.1 server has double RAM than 7.4 server. The "simples" views work right. But when I run the large view the problem appears. The database was migrated with pg_dumpall (8.0.1 version) and restored with psql (8.0.1 version). Please, I need help. Thanks in advance. Indibil postscript: Sorry for my poor english.
On Tue, 2005-02-22 at 14:03, Indibil wrote: > Hi: > > I want migrate to PostgreSQL 8.0.1 for Windows from PostgreSQL 7.4.1, but I > have a big problem. > > In 7.4.1 I have a large view with nested views. It lasts 1 minute and 20 > seconds to finish. In 8.0.1 for Windows it lasts.... I don't know. One > hour, two hour, one day... I always cancel the view because it lasts > forever. > > Shared buffers and work mem in 8.0.1 are greater than those of 7.4.1. So I > think this isn't a memory problem. Furthermore, the 8.0.1 server has double > RAM than 7.4 server. The "simples" views work right. But when I run the > large view the problem appears. > > The database was migrated with pg_dumpall (8.0.1 version) and restored with > psql (8.0.1 version). Have you run vacuum analyze on the new database? Are the postgresql.conf files pretty much the same (except for the renaming and moving around caused by it being an 8.0 .conf file...)
We are noticing on our 7.4.5 system the following files are being written approximately every 5 minutes /tmp/.s.PGSQL.5432 /tmp/.s.PGSQL.5432.lock Anybody know why they are being written? The reason I ask is because we are running on a system that uses a laptop disk drive and requires at least 10 minutes idle time to spindown so it can recalibrate. This did not happen when we were running 7.4.2, so does anybody have any idea what may have changed between 7.4.2 and 7.4.5 to cause this to happen. TIA Chris White
"Chris White (cjwhite)" <cjwhite@cisco.com> writes:
> We are noticing on our 7.4.5 system the following files are being written
> approximately every 5 minutes
> /tmp/.s.PGSQL.5432
> /tmp/.s.PGSQL.5432.lock
> Anybody know why they are being written?
So they don't get eaten by overenthusiastic /tmp cleaner scripts.
> The reason I ask is because we are
> running on a system that uses a laptop disk drive and requires at least 10
> minutes idle time to spindown so it can recalibrate. This did not happen
> when we were running 7.4.2, so does anybody have any idea what may have
> changed between 7.4.2 and 7.4.5 to cause this to happen.
In 7.4 it's done every checkpoint. Perhaps you changed your checkpoint
settings?
In 8.0 it looks like the interval is hardwired to 10 minutes; you'd have
to hack postmaster.c if you don't like it.
regards, tom lane
Thanks for information. We will have to look at the changes we made in
postgresql.conf.
Chris White
-----Original Message-----
From: pgsql-admin-owner@postgresql.org
[mailto:pgsql-admin-owner@postgresql.org] On Behalf Of Tom Lane
Sent: Tuesday, February 22, 2005 3:45 PM
To: cjwhite@cisco.com
Cc: pgsql-admin@postgresql.org
Subject: Re: [ADMIN] 7.4.5 file write issue
"Chris White (cjwhite)" <cjwhite@cisco.com> writes:
> We are noticing on our 7.4.5 system the following files are being
> written approximately every 5 minutes
> /tmp/.s.PGSQL.5432
> /tmp/.s.PGSQL.5432.lock
> Anybody know why they are being written?
So they don't get eaten by overenthusiastic /tmp cleaner scripts.
> The reason I ask is because we are
> running on a system that uses a laptop disk drive and requires at
> least 10 minutes idle time to spindown so it can recalibrate. This did
> not happen when we were running 7.4.2, so does anybody have any idea
> what may have changed between 7.4.2 and 7.4.5 to cause this to happen.
In 7.4 it's done every checkpoint. Perhaps you changed your checkpoint
settings?
In 8.0 it looks like the interval is hardwired to 10 minutes; you'd have to
hack postmaster.c if you don't like it.
regards, tom lane
---------------------------(end of broadcast)---------------------------
TIP 3: if posting/reading through Usenet, please send an appropriate
subscribe-nomail command to majordomo@postgresql.org so that your
message can get through to the mailing list cleanly
Just checked and found that the checkpoint interval is the same between
7.4.2 and 7.4.5. We also brought up a 7.4.2 version and didn't see these
files being written/updated every 5 minutes. In which module are these files
created?
Chris White
-----Original Message-----
From: pgsql-admin-owner@postgresql.org
[mailto:pgsql-admin-owner@postgresql.org] On Behalf Of Chris White (cjwhite)
Sent: Tuesday, February 22, 2005 4:19 PM
To: 'Tom Lane'
Cc: pgsql-admin@postgresql.org
Subject: Re: [ADMIN] 7.4.5 file write issue
Thanks for information. We will have to look at the changes we made in
postgresql.conf.
Chris White
-----Original Message-----
From: pgsql-admin-owner@postgresql.org
[mailto:pgsql-admin-owner@postgresql.org] On Behalf Of Tom Lane
Sent: Tuesday, February 22, 2005 3:45 PM
To: cjwhite@cisco.com
Cc: pgsql-admin@postgresql.org
Subject: Re: [ADMIN] 7.4.5 file write issue
"Chris White (cjwhite)" <cjwhite@cisco.com> writes:
> We are noticing on our 7.4.5 system the following files are being
> written approximately every 5 minutes
> /tmp/.s.PGSQL.5432
> /tmp/.s.PGSQL.5432.lock
> Anybody know why they are being written?
So they don't get eaten by overenthusiastic /tmp cleaner scripts.
> The reason I ask is because we are
> running on a system that uses a laptop disk drive and requires at
> least 10 minutes idle time to spindown so it can recalibrate. This did
> not happen when we were running 7.4.2, so does anybody have any idea
> what may have changed between 7.4.2 and 7.4.5 to cause this to happen.
In 7.4 it's done every checkpoint. Perhaps you changed your checkpoint
settings?
In 8.0 it looks like the interval is hardwired to 10 minutes; you'd have to
hack postmaster.c if you don't like it.
regards, tom lane
---------------------------(end of broadcast)---------------------------
TIP 3: if posting/reading through Usenet, please send an appropriate
subscribe-nomail command to majordomo@postgresql.org so that your
message can get through to the mailing list cleanly
---------------------------(end of broadcast)---------------------------
TIP 8: explain analyze is your friend
"Chris White \(cjwhite\)" <cjwhite@cisco.com> writes:
> Just checked and found that the checkpoint interval is the same between
> 7.4.2 and 7.4.5. We also brought up a 7.4.2 version and didn't see these
> files being written/updated every 5 minutes. In which module are these files
> created?
See the TouchSocketLockFile calls in postmaster.c. I'm pretty sure that
logic did not change between 7.4.2 and 7.4.5 ...
regards, tom lane
> Hi: > > I want migrate to PostgreSQL 8.0.1 for Windows from > PostgreSQL 7.4.1, but I have a big problem. > > In 7.4.1 I have a large view with nested views. It lasts 1 > minute and 20 seconds to finish. In 8.0.1 for Windows it > lasts.... I don't know. One hour, two hour, one day... I > always cancel the view because it lasts forever. > > Shared buffers and work mem in 8.0.1 are greater than those > of 7.4.1. So I think this isn't a memory problem. > Furthermore, the 8.0.1 server has double RAM than 7.4 server. > The "simples" views work right. But when I run the large view > the problem appears. > > The database was migrated with pg_dumpall (8.0.1 version) and > restored with psql (8.0.1 version). > > Please, I need help. Did you run VACUUM ANALYZE on the restored database? //Magnus
I run vacuum analyze and the view WORKS!! A lot of thanks Scott, (one thousand or more ;) Indibil Scott Marlowe wrote: > On Tue, 2005-02-22 at 14:03, Indibil wrote: >> Hi: >> >> I want migrate to PostgreSQL 8.0.1 for Windows from PostgreSQL 7.4.1, but >> I have a big problem. >> >> In 7.4.1 I have a large view with nested views. It lasts 1 minute and 20 >> seconds to finish. In 8.0.1 for Windows it lasts.... I don't know. One >> hour, two hour, one day... I always cancel the view because it lasts >> forever. >> >> Shared buffers and work mem in 8.0.1 are greater than those of 7.4.1. So >> I think this isn't a memory problem. Furthermore, the 8.0.1 server has >> double RAM than 7.4 server. The "simples" views work right. But when I >> run the large view the problem appears. >> >> The database was migrated with pg_dumpall (8.0.1 version) and restored >> with psql (8.0.1 version). > > > Have you run vacuum analyze on the new database? > > Are the postgresql.conf files pretty much the same (except for the > renaming and moving around caused by it being an 8.0 .conf file...) > > ---------------------------(end of broadcast)--------------------------- > TIP 5: Have you checked our extensive FAQ? > > http://www.postgresql.org/docs/faq
On Tue, 22 Feb 2005, Chris White (cjwhite) wrote: CWc> We are noticing on our 7.4.5 system the following files are being written CWc> approximately every 5 minutes CWc> CWc> /tmp/.s.PGSQL.5432 CWc> /tmp/.s.PGSQL.5432.lock CWc> CWc> Anybody know why they are being written? The reason I ask is because we are CWc> running on a system that uses a laptop disk drive and requires at least 10 CWc> minutes idle time to spindown so it can recalibrate. This did not happen CWc> when we were running 7.4.2, so does anybody have any idea what may have CWc> changed between 7.4.2 and 7.4.5 to cause this to happen. What about switching to memory-based /tmp ? Sincerely, D.Marck [DM5020, MCK-RIPE, DM3-RIPN] ------------------------------------------------------------------------ *** Dmitry Morozovsky --- D.Marck --- Wild Woozle --- marck@rinet.ru *** ------------------------------------------------------------------------
Dmitry Morozovsky wrote:
> On Tue, 22 Feb 2005, Chris White (cjwhite) wrote:
>
> CWc> We are noticing on our 7.4.5 system the following files are being written
> CWc> approximately every 5 minutes
> CWc>
> CWc> /tmp/.s.PGSQL.5432
> CWc> /tmp/.s.PGSQL.5432.lock
> CWc>
> CWc> Anybody know why they are being written? The reason I ask is because we are
> CWc> running on a system that uses a laptop disk drive and requires at least 10
> CWc> minutes idle time to spindown so it can recalibrate. This did not happen
> CWc> when we were running 7.4.2, so does anybody have any idea what may have
> CWc> changed between 7.4.2 and 7.4.5 to cause this to happen.
>
> What about switching to memory-based /tmp ?
If you take a look at postmaster.c you will see:
/*
* Touch the socket and lock file at least every ten minutes, to
* ensure that they are not removed by overzealous /tmp-cleaning
* tasks.
*/
now = time(NULL);
if (now - last_touch_time >= 10 * 60)
{
TouchSocketFile();
TouchSocketLockFile();
last_touch_time = now;
}
I would change the "10" to perhaps "60" and try that.
I am now thinking we need to make this "10" a GUC variable for just such
cases. This is not the first request for this.
--
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:
> I am now thinking we need to make this "10" a GUC variable for just such
> cases. This is not the first request for this.
At most it's the second or third request ... and this is hardly the only
repetitive disk hit in Postgres. I can't get excited about it.
regards, tom lane
Tom Lane wrote: > Bruce Momjian <pgman@candle.pha.pa.us> writes: > > I am now thinking we need to make this "10" a GUC variable for just such > > cases. This is not the first request for this. > > At most it's the second or third request ... and this is hardly the only > repetitive disk hit in Postgres. I can't get excited about it. Should we update the 'touch' on the file to once an hour perhaps? -- 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 wrote: > Tom Lane wrote: > > Bruce Momjian <pgman@candle.pha.pa.us> writes: > > > I am now thinking we need to make this "10" a GUC variable for just such > > > cases. This is not the first request for this. > > > > At most it's the second or third request ... and this is hardly the only > > repetitive disk hit in Postgres. I can't get excited about it. > > Should we update the 'touch' on the file to once an hour perhaps? Timer changes to once every hour rather than every 10 minutes. -- 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