[j-nsp] MX104 with full BGP table problems

[Saku Ytti]

On (2014-05-16 13:00 -0500), Brad Fleming wrote:

> We’ve been working with a handful of MX104s on the bench in preparation of putting them into a live network. We started pushing a full BGP table into the device and stumbled across some CPU utilization problems.

What JunOS if 13.3R2 you're probably seeing /var/db/alarm.db recreated every
20, it's 80MB zero filled file.
The file is synchronized over em0 to RE1, causing em congestion and various
other problems, such as loss of connectivity to backup RE1, loss of fan, etc.
And indeed high CPU.

Your em0 traffic should be like <100pps, so the issue is quite obvious if you
graph em0.

I'm using this as workaround:
set system processes alarm-management disable

I dpn't know if it's safe, and JTAC has not yet been able to tell confirm if I
can continue using it. Regardless I'm running it in production in some 12
boxes.
If it's same issue, you might want to refer JTAC to 2014-0430-0067.

I did explain all this when opening the case, asked this week if there is any
progress, and they now told they've found that it's caused by rcp process
causing high I/O load.
I attempted to explain that I don't believe that is the culprit, that is just
symptom of synchronizing the changed file from RE0 to RE1, and perhaps they
should focus on figuring out why the file keeps being recreated.

-- 
  ++ytti