Hello,
Last week we had several freezes of some of our compute nodes. After
investigation we found that we had two VNs with the same CIDR that were
referenced in the same RT. The VMs in theses networks were on the affected
nodes.
We noticed high load on the vrouter-agent and Hold flows. Looking at the
VRF (with rt --dump) the routes to the IPs of the VNs were constantly changing.
We also had a lot of Flow Error messages in the vrouter-agent log. After
some time the node was freezing and a hard-reboot was necessary.
We are still running R1.10~20
Is this kind of bug resolved on R2.20 or > ? I searched issues on
launchpad but I couldn't find any bug related to overlapping and the
agent.
----
Just before crash:
Flow Statistics
---------------
Total Entries --- Total =3D 32016, new =3D 39=20
Active Entries --- Total =3D 31691, new =3D 38=20
Hold Entries --- Total =3D 325, new =3D 1=20
Fwd flow Entries - Total =3D 31613
drop flow Entries - Total =3D 48
NAT flow Entries - Total =3D 30
Rate of change of Active Entries
--------------------------------
current rate =3D 45
Avg setup rate =3D 2068
Avg teardown rate =3D 1096
Rate of change of Flow Entries
------------------------------
current rate =3D 46
After removing the link between the 2 VNs
Flow Statistics
---------------
Total Entries --- Total = 2734, new = -7
Active Entries --- Total = 2734, new = -7
Hold Entries --- Total = 0, new = 0
Fwd flow Entries - Total = 2731
drop flow Entries - Total = 3
NAT flow Entries - Total = 0
Rate of change of Active Entries
--------------------------------
current rate = -9
Avg setup rate = 3689
Avg teardown rate = 0
Rate of change of Flow Entries
------------------------------
current rate = -9
_______________________________________________
Dev mailing list
[email protected]
http://lists.opencontrail.org/mailman/listinfo/dev_lists.opencontrail.org
