SRX

Hi all,

related to my previous post with the subject "High routing memory utilization", I would like to know if there is a command to forcibly empty the buffer poll routing engine before it reaches 100%.

I know the memory should be released automatically but several times it didn't happen causing the device to malfunction.

I've already tried to use "request system storage cleanup" but it only cleans the memory from log files and other such.

Thanks

Would you happen to be using a backup routing engine. If so maybe the problem exists there. Maybe your subsystem is just picky. A recent question I am currently pondering is about windows services. On a machine i7, I have 2 server services that are set to be either disabled or manual at startup. Windows NPS, and routing and remote services. I then after startup make NPS active because the server manager tells me they hang at startup. So should I use routing and remote services? So I'm bringing this up because beyond the total scope of my setup I've got these two issues. If I set up an AD DC it does work. 

I saw another post that says GRES should be setup for reasons I'm not remembering.

Which this next doc might help.

https://www.juniper.net/documentation/us/en/software/junos/high-availability/topics/topic-map/gres-understanding.html

------------------------------
Adrian Aguinaga
B.S.C.M. I.T.T. Tech
(Construction Management)
A.A.S. I.T.T. Tech
(Drafting & Design)
------------------------------

Maybe I missed something from the other post. How did we arrive to looking at buffer pools from high memory utilization? I don't think the two are synonymous.

Besides the graph you shared in the previous post, can you share any other output you've been looking at?

Hi Nikolay,

the graph I posted on my previous post, represents the output of the "show snmp mib get 1.3.6.1.4.1.2636.3.1.13.1.11.9.1.0.0" command. 

The object oid 1.3.6.1.4.1.2636.3.1.13.1.11.9.1.0.0 should represent buffer pool usage.

Compared to when I took the screenshot of the graph, today the output is 60:

============================================

> show snmp mib get 1.3.6.1.4.1.2636.3.1.13.1.11.9.1.0.0
jnxOperatingBuffer.9.1.0.0 = 60

============================================

Despite the OID name, the jnxOperatingBuffer.9.1.0.0 is synonymous with memory memory utilization. 

At any rate, you seem to be looking for a command to just "free memory." Maybe there's some FreeBSD magic out there, but I doubt it. As mentioned in the other thread, you need to look beyond just the overall memory utilization and look at what actually is using so much memory on your device.

Within the JunOS CLI, you can use show system processes extensive to get the memory usage summary at the top.  See https://wiki.freebsd.org/Memory for a brief summary of the memory classes shown there. (Take a look at the External References section, too, some interesting reads there.)

The CLI command will show the processes sorted by CPU usage which may not be very useful in this case. You can drop into a shell with start shell user root and directly call top with option to sort by memory size: top -o size and/or top -o res. That will show you what's actually using so much memory.

And if you want to establish a trend, you have to run these commands periodically (say, once a day) and see what's using more and more memory over time.

Unless you share more information from these observations, I don't think I can help any further.

Regarding "Routing Engine: Memory utilization" is the name of Juniper template for Zabbix.  You are probably right when you say that they are not synonymous