| Author |
Message
|
| fordcam |
 Posted: Fri Feb 02, 2007 10:21 am Post subject: Semaphores on Client Machines |
 |
|
Apprentice
Joined: 28 Mar 2002
Posts: 35
Location: MGIC
|
Our DMZ machines run Solaris, and use the MQ client. I've noticed that we've begun getting FDC files cut on the busiest DMZ machine. All of these FDCs are for probe XY077003, and contain the error "AMQ6119: An internal WebSphere MQ error has occurred ('28 - No space left on device' from semop.)". This error indicates that the kernel's semaphore settings (specifically SEMMNU) are too low.
But the client manual says nothing about kernel configuration. Are there kernel requirements that I ought to be aware of?
FWIW, the server machine handling the connections is pretty lightly used, and isn't reporting any errors at all. And its kernel matches IBM's recommendations. |
|
| Back to top |
|
 |
| Gaya3 |
| Posted: Sat Feb 03, 2007 6:01 pm Post subject: |
|
|
Jedi
Joined: 12 Sep 2006
Posts: 2493
Location: Boston, US
|
|
| Back to top |
|
|
| fordcam |
| Posted: Mon Feb 05, 2007 7:58 am Post subject: |
|
|
Apprentice
Joined: 28 Mar 2002
Posts: 35
Location: MGIC
|
The first link seems to match my problem (the other two aren't client related). I'm surprised that I couldn't find that one; I did all kinds of searches!
I'm also surprised that the Clients manual does not reference kernel parameters anywhere. It seems like that info should be in there, doesn't it? |
|
| Back to top |
|
|
| Vitor |
| Posted: Mon Feb 05, 2007 8:04 am Post subject: |
|
|
Grand High Poobah
Joined: 11 Nov 2005
Posts: 26093
Location: Texas, USA
|
| fordcam wrote: |
I'm also surprised that the Clients manual does not reference kernel parameters anywhere. It seems like that info should be in there, doesn't it? |
It's unknown in my expereince for a client install to require any such changes. My sig refers... 
Perhaps some kind passing IBMer will take up the cause?
_________________
Honesty is the best policy.
Insanity is the best defence. |
|
| Back to top |
|
|
| fordcam |
| Posted: Mon Feb 05, 2007 8:37 am Post subject: |
|
|
Apprentice
Joined: 28 Mar 2002
Posts: 35
Location: MGIC
|
I hate having shared memory and/or semaphore problems. They are generally fatal, and tough to debug. Pretty much all I do is match IBM's settings, and then cross my fingers. But that assumes that they document their settings.  |
|
| Back to top |
|
|
| Vitor |
| Posted: Mon Feb 05, 2007 8:44 am Post subject: |
|
|
Grand High Poobah
Joined: 11 Nov 2005
Posts: 26093
Location: Texas, USA
|
| fordcam wrote: |
| But that assumes that they document their settings. |
They're certainly documented on the server side. I wasn't aware until today the client had requirements!
_________________
Honesty is the best policy.
Insanity is the best defence. |
|
| Back to top |
|
|
| fordcam |
| Posted: Mon Feb 05, 2007 9:02 am Post subject: |
|
|
Apprentice
Joined: 28 Mar 2002
Posts: 35
Location: MGIC
|
| Vitor wrote: |
| fordcam wrote: |
| But that assumes that they document their settings. |
They're certainly documented on the server side. I wasn't aware until today the client had requirements! |
Actually, the 6.0 Quick Beginnings manual left out the SEMMNU requirement, the same setting that we're having problems with on our clients. They later sent out a doc update to include setting SEMMNU to 16384.
Maybe their tech writers have some sort of mental block with this setting.  |
|
| Back to top |
|
|
|
|
