FABNetv4 packet losses from STAR

[Ilya Baldin]

Folks,

I have 3 slices talking to each other over FABNetv4

– Single node ‘storage slice’ on EDC e92a8b7e-4e8e-41c1-a28d-11f31c2a22dc (single node with shared NIC and storage)

– Single node ‘load balancer slice’ on KANS be8cd14d-b24d-4782-bac2-3fd1044b147e (this VM has a Shared NIC and an FPGA)

– 4 node ‘traffic generator slice’ on STAR a34211b3-1f19-4a38-9845-d40c132fb1ff (4 VMs with shared NICs)

They all are able to talk to each other, however the communication between STAR nodes and the other two slices are intermittent with a lot of packet losses. STAR nodes talk to each other without any problems.The storage node and load balancer node can also talk to each other with no problems, so the issue appears to be STAR communicating with the rest of the FABRIC FABNetv4 network.

Here is an example of what I’m seeing when one of the STAR nodes tries to ping the EDC node.

ubuntu@Worker3STAR:~$ ping 10.132.130.2
PING 10.132.130.2 (10.132.130.2) 56(84) bytes of data.
64 bytes from 10.132.130.2: icmp_seq=3 ttl=62 time=2.54 ms
64 bytes from 10.132.130.2: icmp_seq=4 ttl=62 time=2.53 ms
64 bytes from 10.132.130.2: icmp_seq=7 ttl=62 time=2.50 ms
64 bytes from 10.132.130.2: icmp_seq=8 ttl=62 time=2.50 ms
64 bytes from 10.132.130.2: icmp_seq=13 ttl=62 time=2.50 ms

I have all interfaces configured to 9k MTU, but that doesn’t seem to be the issue, since ICMP packets are short. I can keep the STAR slice around if you want to take a look, I’ll create one on a different site in the meantime.

[Ilya Baldin]

I think this is operator error, apologies.

[Author]

Posts