Hi All,
We are having a weird packet lose issue when running VRF+MPLS+OSPF+BGP+BFD between 4 core routers. 2 x 1072 & 2 x 1036.
When pinging outside the VRF there is zero loss but when inside some VRF we see between 1-3% packet loss. The loss seems to correspond with high CPU spikes 95%+ on some cores but we’ve been unable to identify the root cause.
The issue seems to only reside inside the L3VPN and direct traffic and VPLS tunnels seem unaffected.
From the below you can see the CPU load rise and this lasts a couple of seconds the you see a dropped packet.
Loss Occurs
Your assistance is appreciated.
In addition the loss appears to occur around the same time BGP updates.
Every time there is a Drop I execute “/log print where message~“WithdrawnR”” from terminal and there is a update +/- 1 second of the drop.
This is a clear example of Mikrotik being left behind in terms of software development and support.
This is not the type of responce I would expect to a ticket pertaining to a potential bug in their software:
Hello,
Each BGP update/ withdraw triggers routing table recalculation. It also means that after routing table is recalculated table in kernel is also updated which could lead to packet loss during that period.
Best regards,
Maris B.
We run CCR1072’s for a reason and for them not to be able to handle 500mbps of traffic and 600,000 combined routes from 10 BGP peers with standard timers is beyond a joke.
Sure I would expect 1 lost icmp packet out of 10,000 but to loose 1-3 every 100 packets is not by design.
Still trying to get a go / no go from Support.
I hope the original OP still here, I have a prototype network similar to this in works in the labs GNS3 but not quite sure how it will perform in the actual equipment, can somebody can chime in and share their experience can we really use Tiks for this? or I have to look elsewhere? thanks in advance