Hello Benjamin,
T.Rob is right in principle, of course. But I guess, the connections had to go down,
anyway, with these tcp settings. Just in case, what was your tcp_keepcnt? And, did you
check your new tcp parameters took effect immediately? Also, I am not sure if the
change takes effect for the existing connections (never checked that..). What else may
have happened is that the applications were actually alive, although idle, in which
case you do not have much control.. that's why DoS attacks are so nasty when done by
the legitimate apps :-)
Pavel
"Wyatt, T. Rob"
<[EMAIL PROTECTED] To: [EMAIL PROTECTED]
MERICA.COM> cc:
Sent by: MQSeries Subject: Re: many svrconn
instances, as many amqcrsta running
List
<[EMAIL PROTECTED]
C.AT>
06/08/2004 11:05 AM
Please respond to
MQSeries List
Benjamin,
Using QMgr tuning to bring these channels down addresses only the symptom.
Unless you correct the problem at it's root, nothing you do will scale well.
The application MUST properly close it's resources and disconnect from the
QMgr. If you successfully tune the channels, you will probably get through
development and testing but you might not see the real impact until the
application is in Production. On the other hand, if you leave your channels
as they are now, you will have a convenient way to measure the success of
the developers in fixing their code.
-- T.Rob
-----Original Message-----
From: MQSeries List [mailto:[EMAIL PROTECTED] Behalf Of
Benjamin F. Zhou
Sent: Tuesday, June 08, 2004 10:52 AM
To: [EMAIL PROTECTED]
Subject: many svrconn instances, as many amqcrsta running
Hi,
AIX, 5.1, MQ5.3, csd06: we have JMS clients connecting to MQ via client
mode. Initially, the maxchannels got reached quickly, and application
failed at pretty low stress level. After I raised both MaxChannels and
MaxActiveChannels to 400, the test went through pretty well.
However, after each test, I see large number of the svrconn channel and the
same number of amqcrsta running against the same qmgr. There's no sign they
will ever come down, although I set keepAlive=yes and tcp_keepintvl and
tcp_keepidle to 10 seconds.
What else can be done to bring down these orphaned processes?
I saw many postings concerning this or similar problem at mqseries.net ,
just none has found a solution.
Anyone has more idea on this?
best regards,
Benjamin Zhou
Mercedes-Benz USA.
x2474
Instructions for managing your mailing list subscription are provided in
the Listserv General Users Guide available at http://www.lsoft.com
Archive: http://vm.akh-wien.ac.at/MQSeries.archive
Instructions for managing your mailing list subscription are provided in
the Listserv General Users Guide available at http://www.lsoft.com
Archive: http://vm.akh-wien.ac.at/MQSeries.archive
--
This e-mail may contain confidential and/or privileged information. If you are not the
intended recipient (or have received this e-mail in error) please notify the sender
immediately and destroy this e-mail. Any unauthorized copying, disclosure or
distribution of the material in this e-mail is strictly forbidden.
Instructions for managing your mailing list subscription are provided in
the Listserv General Users Guide available at http://www.lsoft.com
Archive: http://vm.akh-wien.ac.at/MQSeries.archive
