[ic] Increasing semaphores arrays

Marty Tennison marty at sediva.com
Wed Jul 16 15:27:38 UTC 2008


Jon Jensen wrote:
> On Tue, 15 Jul 2008, Marty Tennison wrote:
>
>   
>> Has anyone else had problems with Interchange using to many semaphores? 
>> It appears the number of semaphores arrays used by the "interchange" 
>> service is increasing every time the kernel (shared memory/semaphore) 
>> parameter is changed.  We found this by tracking down failed backups.
>>     
>
> The only time I recall running out of semaphores was when running many (> 
> 50) Postgres clusters on the same server. And it was easy to fix.
>
> What mode are you running Interchange in? (Is PreFork enabled?)
>
> Are you using any nonstandard Perl modules?
>
> I don't understand how "failed backups" would lead you anywhere with this. 
> If you can explain that in more detail I might get a better picture of 
> what's going on.
>
> Jon
>
>   
Hi Jon,

Thank you for the reply.

We are running in rpc mode, here are the settings...

Variable  TRAFFIC rpc

PreFork             Yes
StartServers        3
MaxServers          0
MaxRequestsPerChild 100
HouseKeeping        2
PIDcheck            120

perl version is revision 5.0 version 8 subversion 0

I've installed the kitchen sink bundle from cpan as well as a few others 
but none that are experimental. (that I know of)

It appears the semaphore arrays utilized by the interch process 
increased from 121 to 247 and did not leave sufficient resources for the 
CommVault backup process.

ipcs|grep interch |wc -l
247

I increased the share mem/semaphore values to the following:
*** kernel.sem = 506 32000 288 384 ***

After increasing the semaphore value I get this.....

ipcs|grep interch |wc -l
373






More information about the interchange-users mailing list