[j-nsp] MX80 BGP performance after reboot

[Sebastian Wiesinger]

* Sebastian Wiesinger <juniper-nsp at ml.karotte.org> [2013-02-14 08:06]:
> * Stacy W. Smith <stacy at acm.org> [2013-02-12 01:18]:
> 
> > Do the KRT error messages go away if you unconfigure sampling? Any
> > change in the KRT installation time with sampling turned off?
> 
> I'll test that. I assume I will need to completely disable the
> sampling instance?

So,

I just tested this after talking to JTAC. Just for reference:

I had ~70k routes from 40 peers that I deactivated. I then turned them
up again and measured with inline-jflow disabled and enabled.

With inline-jflow ON: around 10-12 minutes until KRT settles
With inline-jflow OFF: around 2 minutes until KRT settles

To build a full table (after reboot) it's even longer of course.

This produces very nice loops while it's settling. I can disable
inline-jflow when doing upgrades/planned maintenance but I need it the
rest of the time. So... out of luck for unplanned BGP session resets.

I'm waiting for JTAC on how to proceed.

Regards

Sebastian

-- 
GPG Key: 0x93A0B9CE (F4F6 B1A3 866B 26E9 450A  9D82 58A2 D94A 93A0 B9CE)
'Are you Death?' ... IT'S THE SCYTHE, ISN'T IT? PEOPLE ALWAYS NOTICE THE SCYTHE.
            -- Terry Pratchett, The Fifth Elephant