|
Neohapsis is currently accepting applications for employment. For more information, please visit our website www.neohapsis.com or email hr@neohapsis.com |
Subject: RE: 1012 noops in a row cause slow responses from Postfix.
From: Mark Hoffman (mark.hoffman
wallst.com)Date: Fri Jun 23 2000 - 14:50:35 CDT
- Next message: Wietse Venema: "Re: 1012 noops in a row cause slow responses from Postfix."
- Previous message: Marco d'Itri: "Re: Content-Length (was Re: Banner)"
- Maybe in reply to: Mark Hoffman: "1012 noops in a row cause slow responses from Postfix."
- Next in thread: Wietse Venema: "Re: 1012 noops in a row cause slow responses from Postfix."
- Maybe reply: Mark Hoffman: "RE: 1012 noops in a row cause slow responses from Postfix."
- Reply: Wietse Venema: "Re: 1012 noops in a row cause slow responses from Postfix."
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
I set the service to noop every 2 seconds to reproduce the problem.
I guess the real problem is that even after a successful delivery, the
response is still slow. We are seeing this behavior even while sending email
in the middle of the 1012 noops. i.e. If we send 500 noops, then an email,
Then more noops, response gets slow at ~1030 noops total. So it seems the
successful email doesn't reset the session record, when a constant
connection is maintained.
In our production environment, we see this behavior after ~ 3 days. Granted,
NT usually has to be rebooted more often than that, but I've finally gotten
my service to run for a couple of weeks at a stretch. And, because of the
load on the gateways between the NT servers and the Postfix servers, I want
to noop to make sure the connection is still there before sending mail.
Thanks,
Mark
-----Original Message-----
From: wietse
porcupine.org [mailto:wietse
porcupine.org]
Sent: Friday, June 23, 2000 3:02 PM
To: postfix-users
postfix.org
Subject: Re: 1012 noops in a row cause slow responses from Postfix.
Mark Hoffman:
> My NT service (ick.) is maintaining a 24/7 connection to a bank of Postfix
> servers. The service is noop'ing every 2 seconds to maintain the
connection,
> and to check that the connection is alive before sending out mail. When
the
> commands sent to Postfix hit 1012, we get very slow response times. Has
> anyone else seen this? This has been a recent development in Postfix as
the
> version I used in the middle of last year didn't exhibit this feature. Any
> help would be appreciated.
The limit is by default:
smtpd_junk_command_limit = 1000
The SMTP server maintains a record of each session. The record is
reset after each successful delivery.
The limit on the number of junk commands prevents the session record
from consuming unbounded amounts of memory.
Sending NOOP every two seconds makes no sense. By default, Postfix
times out an idle connection after 300 seconds:
smtpd_timeout = 300
I suggest you fix the NT client software.
Wietse
- Next message: Wietse Venema: "Re: 1012 noops in a row cause slow responses from Postfix."
- Previous message: Marco d'Itri: "Re: Content-Length (was Re: Banner)"
- Maybe in reply to: Mark Hoffman: "1012 noops in a row cause slow responses from Postfix."
- Next in thread: Wietse Venema: "Re: 1012 noops in a row cause slow responses from Postfix."
- Maybe reply: Mark Hoffman: "RE: 1012 noops in a row cause slow responses from Postfix."
- Reply: Wietse Venema: "Re: 1012 noops in a row cause slow responses from Postfix."
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]