| Author |
Message
|
| nathanw |
 Posted: Fri Mar 03, 2006 7:13 am Post subject: Show Stopper - Channel Limits Connections and Problems |
 |
|
Knight
Joined: 14 Jul 2004
Posts: 550
|
I have a win xp running 5.3 CSD 7 environment (512 meg RAM, 80 gb HD)
10 Queues Managers configured on 10 separate ports (1421 - 1430)
They all need to connect to a queuemanager on a separate machine (SLES 9)
So 10 - 1 connection Queuemanager wise (as far as I can see No problems there)
Problems encountered
On a LAN connection I can only get 3 Sender Channels to run between machines even when the network card is forced to run at 100mbps on full duplex
All other Channels sit at Binding and do not connect
There is no firewall as it is an internal secure network
There is no data being passsed at this point just connections
Any ideas
Thanks
Last edited by nathanw on Fri Mar 03, 2006 7:25 am; edited 1 time in total |
|
| Back to top |
|
 |
| mvic |
| Posted: Fri Mar 03, 2006 7:23 am Post subject: Re: Show Stopper - Channel Limits Connections and Problems |
|
|
Jedi
Joined: 09 Mar 2004
Posts: 2080
|
| nathanw wrote: |
| I have a win xp running 5.3 CSD 7 environment |
OK this is going to be point-by-point. (If anything looks like a discussion separate from the channel issue, perhaps it's best to start a new thread for that thing). First point is that CSD07 is old now. We're up to CSD12 now, and it looks good.
| Quote: |
| (512 meg RAM, 80 gb HD) / 10 Queues Managers configured on 10 separate ports (1421 - 1430) |
I hope you don't have to handle many large or even medium-sized messages. This spec is pretty tiny for 10 queue managers.
| Quote: |
| All other Channels sit at Binding and do not connect / There is no data being passsed at this point just connections / Any ideas |
Are any of these channels triggered? Try passing some test data, and see what happens. |
|
| Back to top |
|
|
| nathanw |
| Posted: Fri Mar 03, 2006 7:34 am Post subject: |
|
|
Knight
Joined: 14 Jul 2004
Posts: 550
|
| mvic wrote: |
| First point is that CSD07 is old now. We're up to CSD12 now, and it looks good. |
I agree and will be updating the CSD BUT will this improve any performance
| mvic wrote: |
| I hope you don't have to handle many large or even medium-sized messages. This spec is pretty tiny for 10 queue managers. |
I agree but is all I have to work with at the moment, would you recomend I try a lesser amount? as each has it's own listener on each port would fewer queue managers and therefore fewer listeners improve the situation? - Your thoughts appreciated
Message size will be small but a large number of them
This will simulate the real world except that there will be multiple machines with single queuemanagers all talking to the single server.
| mvic wrote: |
| Are any of these channels triggered? Try passing some test data, and see what happens. |
All Channels have triggers associated with them. I have not tried passing any daya through as yet as all I want to do is establish comms across all 10 (refer above)
Maybe I should aim for fewer Numbers |
|
| Back to top |
|
|
| jefflowrey |
| Posted: Fri Mar 03, 2006 7:36 am Post subject: |
|
|
Grand Poobah
Joined: 16 Oct 2002
Posts: 19981
|
So you have manually started each channel on each queue manager and only some of them have successfully started, the rest went into bindings.
Wait until those channels go into retry, and then see what errors occur on each side.
_________________
I am *not* the model of the modern major general. |
|
| Back to top |
|
|
| nathanw |
| Posted: Fri Mar 03, 2006 7:45 am Post subject: |
|
|
Knight
Joined: 14 Jul 2004
Posts: 550
|
| jefflowrey wrote: |
So you have manually started each channel on each queue manager and only some of them have successfully started, the rest went into bindings.
Wait until those channels go into retry, and then see what errors occur on each side. |
AHHH OK will do
I suspect it is a network problem but have yet to confirm
Will check though |
|
| Back to top |
|
|
| mvic |
| Posted: Fri Mar 03, 2006 7:58 am Post subject: |
|
|
Jedi
Joined: 09 Mar 2004
Posts: 2080
|
| nathanw wrote: |
| I agree and will be updating the CSD BUT will this improve any performance |
Almost certainly not. But CSDs fix problems, including perhaps the one you're seeing here?
| nathanw wrote: |
| mvic wrote: |
| I hope you don't have to handle many large or even medium-sized messages. This spec is pretty tiny for 10 queue managers. |
I agree but is all I have to work with at the moment, would you recomend I try a lesser amount? as each has it's own listener on each port would fewer queue managers and therefore fewer listeners improve the situation? - Your thoughts appreciated |
Either you need a listener or you don't. Either you need a queue manager or you don't (OK I know this is sometimes a matter of opinion). But seeing as you are intending to simulate a real production scenario, this is not a "normal" example.
In any case, I don't see that the number of queue managers or listeners (10 of each) would itself cause problems with channels. There must be another explanation, IMHO.
For what it's worth, given a free choice, I would probably restrict such a small machine to running just one or two queue managers, especially if it/they had to handle high throughput.
| Quote: |
Message size will be small but a large number of them
This will simulate the real world except that there will be multiple machines with single queuemanagers all talking to the single server. |
OK - just be aware you won't be able to simulate the performance characteristics of your intended production environment, for obvious reasons.
| Quote: |
All Channels have triggers associated with them. I have not tried passing any daya through as yet as all I want to do is establish comms across all 10 (refer above)
Maybe I should aim for fewer Numbers |
If this is not production, then I don't see a problem necessarily. But remember you won't be able to simulate large throughputs that larger machines are capable of.
If channels are triggered, try sending messages through them and see what happens. |
|
| Back to top |
|
|
| nathanw |
| Posted: Sat Mar 04, 2006 2:21 am Post subject: SOLVED - Show Stopper - Channel Limits Connections and Probl |
|
|
Knight
Joined: 14 Jul 2004
Posts: 550
|
Well I deciede last night to try a differnt tack
I used the Apollo 13 approach just starting things up in the right order
Well in the end I managed 8 queuemangers all running with channels in 2 way comms and passing data through
This will make my life alot easier at least.
Also think the network was being hampered elsewhere.
We shall see
BTW thanks for your help and suggestiions guys. |
|
| Back to top |
|
|
| BBM |
| Posted: Mon Mar 06, 2006 2:08 pm Post subject: |
|
|
Master
Joined: 10 Nov 2005
Posts: 217
Location: London, UK
|
|
| Back to top |
|
|
| nathanw |
| Posted: Tue Mar 07, 2006 2:57 am Post subject: |
|
|
Knight
Joined: 14 Jul 2004
Posts: 550
|
Very similar
I was already starting them under STRMQM anyway as they were being created, edited and started by a script
for testing purposes I casn live with 8 to a machine
would be interesting to actually find a problem one day when IBM dont blame MS and vice versa |
|
| Back to top |
|
|
| mvic |
| Posted: Tue Mar 07, 2006 3:10 am Post subject: |
|
|
Jedi
Joined: 09 Mar 2004
Posts: 2080
|
| nathanw wrote: |
| for testing purposes I casn live with 8 to a machine |
IMHO it's an unusual requirement to run more than a couple of queue managers on one Windows machine.
BUT, even so...
I used Google and found that Windows desktop heap issues (this is what you're running into here, right?) are not peculiar to MQ, or to IBM software running on Windows - it appears instead to be a known limitation in Windows when using the default desktop heap parameters.
See http://www.google.co.uk/search?q=increase+desktop+heap
(From the Google search) Microsoft themselves document a workaround : http://support.microsoft.com/default.aspx?kbid=126962
EDIT: add link to KB article |
|
| Back to top |
|
|
| nathanw |
| Posted: Tue Mar 07, 2006 3:46 am Post subject: |
|
|
Knight
Joined: 14 Jul 2004
Posts: 550
|
msny thanks
have checked and edited we will see |
|
| Back to top |
|
|
|
|
