| Author |
Message
|
| KAKEZ |
 Posted: Fri Jan 31, 2003 5:34 am Post subject: How many simultaneous requests can inetd handle? |
 |
|
Centurion
Joined: 10 Oct 2002
Posts: 117
|
Hi,
some wonderings about inetd:
general:
- how many simultaneous requests can inetd handle?
- is there a limit?
- can we specify a max number of requests inetd can handle - this to eventually install several inted listeners on differents ports
for client connections:
- defining group of Qmgrs in the channel definition table AMQCLCHL.TAB
allowing connection tries on different client channels it is possible to think about different client channels on the same Qmgr but on different port numbers -> several inetd on different ports
thanks for feeling about that?
Jack |
|
| Back to top |
|
 |
| bduncan |
| Posted: Fri Jan 31, 2003 9:31 am Post subject: |
|
|
Padawan
Joined: 11 Apr 2001
Posts: 1554
Location: Silicon Valley
|
I am fairly certain that the total number of connections is limited only by the number of ports available. However, the rate at which connections are created by inetd does appear to have an upper limit. Look at the following link. It explains the problem with respect to IMAP rather than MQSeries, but the same limitations apply. Notice the followup to the post tells you how to modify this limit fairly easily on Solaris. I guess the best bet is to do 'man inetd' on your system to find out the particulars for your platform:
http://www.washington.edu/imap/listarch/2001/msg00013.html
_________________
Brandon Duncan
IBM Certified MQSeries Specialist
MQSeries.net forum moderator |
|
| Back to top |
|
|
| EddieA |
| Posted: Fri Jan 31, 2003 9:49 am Post subject: |
|
|
Jedi
Joined: 28 Jun 2001
Posts: 2453
Location: Los Angeles
|
MQ 5.3 has introduced a new concept of 'pooling' for the listeners which allows a higher number of connections.
It's there, by default, if you use runmqlsr.
I get the impression that this is the intended direction, as nearly all the references to inetd have been pulled from the manuals.
Cheers,
_________________
Eddie Atherton
IBM Certified Solution Developer - WebSphere Message Broker V6.1
IBM Certified Solution Developer - WebSphere Message Broker V7.0 |
|
| Back to top |
|
|
| bower5932 |
| Posted: Fri Jan 31, 2003 9:53 am Post subject: |
|
|
Jedi Knight
Joined: 27 Aug 2001
Posts: 3023
Location: Dallas, TX, USA
|
| runmqlsr is definitely the intended direction. |
|
| Back to top |
|
|
| KAKEZ |
| Posted: Sun Feb 02, 2003 10:51 pm Post subject: |
|
|
Centurion
Joined: 10 Oct 2002
Posts: 117
|
Hi guys,
thanks to all for interesting iinformation,
but other question:
- i have heard that inetd is more robust than runmqlsr so if runmqlsr comes down all the connections it handles are lost
- with inetd we have one amqcrsta for each connection witch is much more sure - only one connection lost in case of fail of amqcrsta
so what actually do you recommend - runmqlsr or inetd ?
Jack |
|
| Back to top |
|
|
| bower5932 |
| Posted: Mon Feb 03, 2003 5:37 am Post subject: |
|
|
Jedi Knight
Joined: 27 Aug 2001
Posts: 3023
Location: Dallas, TX, USA
|
With WebSphere MQ 5.3, the intended direction is to steer people towards runmqlsr. It is better designed to handle a large number of connections because of the pooling that is based on threads rather than amqcrsta which is process based.
Besides, runmqlsr won't come down.....  |
|
| Back to top |
|
|
| bduncan |
| Posted: Mon Feb 03, 2003 1:26 pm Post subject: |
|
|
Padawan
Joined: 11 Apr 2001
Posts: 1554
Location: Silicon Valley
|
I believe on UNIX you can start runmqlsr via inittab and set the respawn parameter, so that if it does "come down" the init process will bring it right back up. So on the surface it will behave just like starting via inetd.
_________________
Brandon Duncan
IBM Certified MQSeries Specialist
MQSeries.net forum moderator |
|
| Back to top |
|
|
| davidmadison |
| Posted: Wed Feb 12, 2003 8:57 am Post subject: inetd requests |
|
|
Newbie
Joined: 07 Feb 2003
Posts: 2
|
I find it odd that the direction for 5.3 is to use runmqlsr, given that in earlier versions inetd was recommended, unless using MS Windows.
Also, if you are going to use runmqlsr, get the patch.
5.2: U482979
5.3: U484025
otherwise, the runmqlsr process may die if more than 252 connections are made (3 for stdio + 252 == 255)
David |
|
| Back to top |
|
|
| vmcgloin |
| Posted: Thu Feb 13, 2003 1:30 am Post subject: |
|
|
Knight
Joined: 04 Apr 2002
Posts: 560
Location: Scotland
|
David,
Please can you post the details of these patches - where to find them & descriptions? We have had a PMR open with IBM for weeks now for exactly the problem you describe & they have not come up with any solution.
Thanks!
Vicky |
|
| Back to top |
|
|
| bower5932 |
| Posted: Thu Feb 13, 2003 5:09 pm Post subject: |
|
|
Jedi Knight
Joined: 27 Aug 2001
Posts: 3023
Location: Dallas, TX, USA
|
I'm at the Transaction and Messaging Conference, and one of the presentations pointed out that with WebSphere MQ 5.3 the preferred listener is runmqlsr. It now has connection pooling and can handle more connections.
This is a definite change from what was recommended with earlier releases. |
|
| Back to top |
|
|
|
|
