Hi Forum,
I have lodged a support ticket but thought I’d put this out there:
I have a CCR1036-8G-2S+ running in production.
It was installed with 6.13 running, where the problem was apparant. We’ve since upgraded to 6.19 and the problem continues.
I’m not willing to run release candidate on production - and I can’t reproduce the issue to set up in a lab.
The Problem:
Every so often (it is sporadic), all 36 cores will jump to 100% usage, all streams will halt (including management/winbox) for about 1-3 seconds.
Unfortunately winbox freezes during this but when watching the profiler - when its trying to catch up the ‘networking’ process appears to be the one chewing up extra time.
BGP peers stay established so it doesn’t appear to be a large BGP import/export causing the problem. Logs don’t seem to indicate any wrong doing.
For TCP streams its almost unnoticeable. I notice it in SSH session and the like, but the real problem is UDP streams. We run a lot of VoIP and the audio drops completely while the issue occurs.
We’ve tried dropping the load on the box but the spikes still happen.
Has anyone else experienced such an issue?
To explain some of the config specifics:
~20 filter rules
~20 packet marking rules
~10 Simple queues
Bridge interface contains ~10 VLANs that almost all data travels through.
802.3AD bonding interface is member of the bridge interface.
2 1gbit ethernet links are members of the bonding interface.
The bonding interface was member of a bridge as there was to be another bonding interface to a different switch also member of the bridge, for redundancy.
There is also 1 SFP+ module (10gbit) that runs around 100mbit of traffic at peak.
Thanks