[AGENT++] Memory issue with Informs

Mayank Thakore mayankt at huawei.com
Fri Sep 28 22:24:44 CEST 2007


Found the problem ! In Solaris, the call the free() is not releasing memory
back to the system. So once the alarms pile up the memory usage goes up, but
even when they are sent and deleted the memory doesn't come down. Though
later the memory stabilizes around the max required at any given time.

Sorry for the mails. And thanks for the suggestions. I think we will use the
suggestions anyway. 

Regards
Mayank Thakore
Huawei Technologies
Mobile: +91-988-665-9544

 
****************************************************************************
****************************
 This e-mail and attachments contain confidential information from HUAWEI,
which is intended only for the person or entity whose address is listed
above. Any use of the information contained herein in any way (including,
but not limited to, total or partial disclosure, reproduction, or
dissemination) by persons other than the intended recipient's) is
prohibited. If you receive this e-mail in error, please notify the sender by
phone or email immediately and delete it!

-----Original Message-----
From: Frank Fock [mailto:fock at agentpp.com] 
Sent: Saturday, September 29, 2007 00-05
To: mayankt at huawei.com
Cc: Bob.Natale at appliedsnmp.com; agentpp at agentpp.org
Subject: Re: [AGENT++] Memory issue with Informs

Hi Mayank,

Like Bob, I guess that memory consumption is caused
by massive concurrent waiting on INFORM responses.

If your system is sending that much traps/informs
you should think about redesigning towards some
sort of alarm aggregation/filter. Otherwise,
also the management application will have problems
to process the trap/inform flood.

Best regards,
Frank


Mayank Thakore wrote:
> Hi Bob,
> 
> We are using the sync/blocking call:
> 
> 	virtual int inform(Pdu &pdu, SnmpTarget &target);
> 
> If it fails, we wait for user configured seconds and try again with the
same
> Pdu and SnmpTarget. And we keep trying until our agent is shutdown or the
> inform is successfully sent.
> 
> Even when informs are being sent successfully (not necessarily in the
first
> try, we set the retry to 3 attempts) the memory rises.
> 
> I will try testing by changing the retry count and not using our own retry
> code. Meanwhile can you give any other pointers?
> 
> If we use trap() instead of inform() the problem disappears.
> 
> Thanks for the help.
> 
> 
> Regards
> Mayank Thakore
> Huawei Technologies
> Mobile: +91-988-665-9544
>  
>
****************************************************************************
> ****************************
>  This e-mail and attachments contain confidential information from HUAWEI,
> which is intended only for the person or entity whose address is listed
> above. Any use of the information contained herein in any way (including,
> but not limited to, total or partial disclosure, reproduction, or
> dissemination) by persons other than the intended recipient's) is
> prohibited. If you receive this e-mail in error, please notify the sender
by
> phone or email immediately and delete it!
> ________________________________________
> From: Bob Natale [mailto:bob.natale at appliedsnmp.com] 
> Sent: Friday, September 28, 2007 22-28
> To: mayankt at huawei.com
> Cc: agentpp at agentpp.org
> Subject: Re: [AGENT++] Memory issue with Informs
> 
> HI Mayank,
>  Are you seeing Responses to the Informs?
> I would guess that some component is holding state for each Inform pending
> Response processing -- emphasis on "guess"! :)
> Cheers,
> BobN
> 
> -----Original Message-----
> From: Mayank Thakore [mailto:mayankt at huawei.com]
> Sent: Friday, September 28, 2007 07:02 AM
> To: agentpp at agentpp.org
> Subject: [AGENT++] Memory issue with Informs
> 
> Hi,
> 
> Our SNMP++ version is 2.6
> Our Agent++ version is 3.5.13a
> 
> We are facing an issue with V2 informs. When we send the inform memory
usage
> rises rapidly as we send more and more informs. We tested our whole code
and
> found no leaks. Then we finally replaced the inform call with trap call.
The
> problem was no longer seen. So we suspect that V2 informs are using up
> memory somewhere.
> 
> We cannot confirm it is a leak since Valgrind did not report any leaks.
Can
> someone please tell if there is an issue which will cause memory to be
used
> up rapidly when sending many V2 informs?
> 
> Much thanks for the help.
> 
> Regards
> Mayank Thakore
> Huawei Technologies
> Mobile: +91-988-665-9544
> 
> p.s. we checked the Agent++ change log and saw several leaks fixed since
> release 3.5.13a but cannot figure out if any impact us in this case. And
we
> don't want to upgrade unless we have to.
>  
>
****************************************************************************
> ****************************
>  This e-mail and attachments contain confidential information from HUAWEI,
> which is intended only for the person or entity whose address is listed
> above. Any use of the information contained herein in any way (including,
> but not limited to, total or partial disclosure, reproduction, or
> dissemination) by persons other than the intended recipient's) is
> prohibited. If you receive this e-mail in error, please notify the sender
by
> phone or email immediately and delete it!
> 
> 
> 
> _______________________________________________
> AGENTPP mailing list
> AGENTPP at agentpp.org
> http://lists.agentpp.org/mailman/listinfo/agentpp
> 
> 
> _______________________________________________
> AGENTPP mailing list
> AGENTPP at agentpp.org
> http://lists.agentpp.org/mailman/listinfo/agentpp

-- 
AGENT++
http://www.agentpp.com
http://www.mibexplorer.com
http://www.mibdesigner.com






More information about the AGENTPP mailing list