[j-nsp] Monitoring memory usage via SNMP

[Wilkes, Nathan]

>From what I understand, Juniper modified the calculation to more accurately represent the amount of memory free/in use.

It used to be calculated like so:
	Memory utilization % = (Total - Free - Inactive)/Total

In 7.6, it is calculated like this:
	Memory utilization % = (Total - Free - cached)/Total

The amount of memory in use did not change (except for the obligatory changes due to the new code version) -- only the calculation.

The problem with the earlier calculation was that Inactive Memory is not easily accessible by the kernel.  The kernel assigns memory to various processes, like RPD for example.  The portion of the assigned memory that is not used by the process is marked "Inact" so that the kernel can pull from it if necessary after using all of the Free and Cache memory.  The kernel would still have to run a paging routine to get the memory back (or swap, etc.), which is obviously more intensive than just pulling from free/cache memory.

We just ended up increasing our monitored thresholds for alerting.  As long as the router isn't swapping or using flash (which it does when it gets up near 100%), it runs just fine.

- Nathan

-----Original Message-----
From: juniper-nsp-bounces at puck.nether.net
[mailto:juniper-nsp-bounces at puck.nether.net]On Behalf Of Jee Kay
Sent: Saturday, November 25, 2006 10:27 AM
To: Lars Erik Gullerud
Cc: juniper-nsp at puck.nether.net
Subject: Re: [j-nsp] Monitoring memory usage via SNMP


On 21/04/06, Lars Erik Gullerud <lerik at nolink.net> wrote:
> If you just need to look at the memory usage on the RE, you can poll the
> jnxOperatingBuffer (1.3.6.1.4.1.2636.3.1.13.1.11) and, if necessary, the
> jnxOperatingHeap (1.3.6.1.4.1.2636.3.1.13.1.12) tables, both from the
> Juniper Chassis-MIB. We do this from InfoVista to monitor general memory

This has been working fantastically for the last 9 or so months, but
after an upgrade to JunOS 7.6 it appears this variable is now
calculated differently... where before it was Active/Total, it now
appears to be returning (Total-Inactive)/Total, resulting in my
routers now reporting 83% utilisation instead of 17% as they were
before.

Has anyone else seen this / is there now a better OID to monitor? I'm
a little wary of setting my notification limits above 85% for obvious
reasons.

Thanks,
Ras
_______________________________________________
juniper-nsp mailing list juniper-nsp at puck.nether.net
https://puck.nether.net/mailman/listinfo/juniper-nsp