RT not multitasking well

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Greetings…

We switched over to RT2 a few months back, and things have been
overall well. However, we’re having a little bottlenecking issue –
it only seems to be doing one SQL query at a time, even if there’s
multiple folks hammering it via the web interface. This gets really
nasty, as we go through a LOT of tickets and have been known to do
bulk operations with over a thousand tickets at a time :slight_smile:

We’re running RT 2.0.13 with PostgreSQL 7.1.3 on a RedHat 7.2 box,
and using FastCGI… most modules and other stuff are current as of
early June, but I can dig up specific version numbers if it matters
much :wink: Config snippets can be dredged up as well.

Any help, pointers at docs I’ve missed, or suggestions of good beers
would be appreciated. Thanks! -rt


Ryan Tucker rtucker@netacc.net
GPG: 99A27DE7 AIM: HoopyCat ARS: N0PBS
Network Operations Manager, NetAccess, Inc.
http://www.netacc.net/~rtucker/ - (585)419-8252
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.0.7 (Darwin)

iD8DBQE9RdGRDu//cZmifecRAiylAKDiFpMAYgq+fBp7rD+vz2qmsdiQzQCgw4rI
ny6rx83zTCDCz6LXrg9AEdw=
=tBf6
-----END PGP SIGNATURE-----

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Greetings…

We switched over to RT2 a few months back, and things have been
overall well. However, we’re having a little bottlenecking issue –
it only seems to be doing one SQL query at a time, even if there’s
multiple folks hammering it via the web interface. This gets really
nasty, as we go through a LOT of tickets and have been known to do
bulk operations with over a thousand tickets at a time :slight_smile:

Have you vacuumed the tables in the RT database lately? Ever?

If not, you should :slight_smile:

- Justin

Justin Hawkins
Internode Professional Access

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1On Monday, July 29, 2002, at 08:09 , Justin Hawkins wrote:

Have you vacuumed the tables in the RT database lately? Ever?

If not, you should :slight_smile:

Nightly :slight_smile: That one became handy about, oh, a week in… :slight_smile: -rt


Ryan Tucker rtucker@netacc.net
GPG: 99A27DE7 AIM: HoopyCat ARS: N0PBS
Network Operations Manager, NetAccess, Inc.
http://www.netacc.net/~rtucker/ - (585)419-8252
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.0.7 (Darwin)

iD8DBQE9ReLVDu//cZmifecRAqVzAJ4nUlt0korTT1+127vQ5QvRzBDp+QCeMqRk
wRaukHZxjmrf5EpYjJfwRmA=
=CmsW
-----END PGP SIGNATURE-----

Have you vacuumed the tables in the RT database lately? Ever?

If not, you should :slight_smile:

Nightly :slight_smile: That one became handy about, oh, a week in… :slight_smile: -rt

Yep OK :slight_smile:

Does something along the lines of ‘ps axwwww | grep postmaster’ on the
database server show anything interesting during attempted concurrent
transactions?

There should be a postmaster per httpd - at least that’s what I have. Do
you have enough httpd child processes around to service requests?

Just stabbing in the dark here :wink:

- Justin

Justin Hawkins
Internode Professional Access

Hmm, I have precisely one postmaster and 13 httpds… and I do
think httpd is more or less spawning OK, because there’s significant
other content on that box which is fine.

Interesting… -rt

Indeed:

56489 ?? I 1:33.09 postmaster: rt_user rt2 192.168.0.1 idle (postgres)
63614 ?? I 1:33.43 postmaster: rt_user rt2 192.168.0.1 idle (postgres)
96002 ?? I 0:59.95 postmaster: rt_user rt2 192.168.0.1 idle (postgres)
96381 ?? I 1:55.81 postmaster: rt_user rt2 192.168.0.1 idle (postgres)
97346 ?? I 1:26.07 postmaster: rt_user rt2 192.168.0.1 idle (postgres)
97531 ?? I 1:57.92 postmaster: rt_user rt2 192.168.0.1 idle (postgres)
98160 ?? I 1:24.46 postmaster: rt_user rt2 192.168.0.1 idle (postgres)

Which corresponds to:

25704 ?? I 1:33.06 /usr/local/sbin/httpd
35910 ?? I 2:07.36 /usr/local/sbin/httpd
37299 ?? I 2:38.06 /usr/local/sbin/httpd
43995 ?? I 3:25.80 /usr/local/sbin/httpd
58467 ?? I 3:09.62 /usr/local/sbin/httpd
58469 ?? I 2:20.23 /usr/local/sbin/httpd
58470 ?? I 2:18.06 /usr/local/sbin/httpd

My only difference is that the web server and postgres are on different
machines.

Can send you my relevant httpd.conf lines if you like, with the caveat
that I don’t really understand it, I just pasted it in :slight_smile:

- Justin

Justin Hawkins
Internode Professional Access

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1On Tuesday, July 30, 2002, at 02:00 , Justin Hawkins wrote:

Hmm, I have precisely one postmaster and 13 httpds… and I do
think httpd is more or less spawning OK, because there’s significant
other content on that box which is fine.

Interesting… -rt

Indeed:

[snip]

My only difference is that the web server and postgres are on different
machines.

Can send you my relevant httpd.conf lines if you like, with the caveat
that I don’t really understand it, I just pasted it in :slight_smile:

It’s worth a shot… I do think there’s a good chance I missed
something in there, because what you’re seeing is quite happy
behavior.

Thanks :slight_smile: -rt


Ryan Tucker rtucker@netacc.net
GPG: 99A27DE7 AIM: HoopyCat ARS: N0PBS
Network Operations Manager, NetAccess, Inc.
http://www.netacc.net/~rtucker/ - (585)419-8252
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.0.7 (Darwin)

iD8DBQE9Ri11Du//cZmifecRAgFiAJ9BnricqEEwr+iS9CoSXhKUAbwl4QCfSIdS
4pcqLOnnup5skkETyUQEqO0=
=ChgB
-----END PGP SIGNATURE-----

Have you vacuumed the tables in the RT database lately? Ever?

Vacuumed?

Is there a script (script, not scrip :slight_smile: available to clean up old stuff
out of the database?

Steve Sobol, CTO JustThe.net LLC, Mentor On The Lake, OH
Temporary Office Phone: 786.666.5357 (Due to incompetence on the part
of XO Communications, my local phone line is down, and therefore, my
tollfree numbers don’t work either.)

Have you vacuumed the tables in the RT database lately? Ever?

Vacuumed?

Is there a script (script, not scrip :slight_smile: available to clean up old stuff
out of the database?

VACUUM is a Postgres command that cleans old Postgres stuff out of the
database – fills up holes and so forth. It’s not RT-specific. (What
sort of “old stuff” were you hoping to remove? There’s always that
problem of referential integrity…)

-Rich

Rich Lafferty --------------±----------------------------------------------
Ottawa, Ontario, Canada | Save the Pacific Northwest Tree Octopus!
http://www.lafferty.ca/ | http://zapatopi.net/treeoctopus.html
rich@lafferty.ca -----------±----------------------------------------------

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1On Tuesday, July 30, 2002, at 02:08 , Ryan Tucker wrote:

It’s worth a shot… I do think there’s a good chance I missed
something in there, because what you’re seeing is quite happy
behavior.

As an update, it looks like FastCGI was only spawning a single
postgres process, therefore slowing things down. Disabling that,
which was enabled and then forgotten about while traffic increased
drastically, has sped things up noticeably.

Thanks all :slight_smile: -rt


Ryan Tucker rtucker@netacc.net
GPG: 99A27DE7 AIM: HoopyCat ARS: N0PBS
Network Operations Manager, NetAccess, Inc.
http://www.netacc.net/~rtucker/ - (585)419-8252
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.0.7 (Darwin)

iD8DBQE9Sv+1Du//cZmifecRAnp6AKC15bKBt6wtKrneUiT05irq5MjydQCg8Maw
HDWOCHrDwYLYrKoMW18r/9U=
=SuwW
-----END PGP SIGNATURE-----