** Description changed: - Ubuntu 20.04 LTS - dpdk 19.11.7-0ubuntu0.20.04.1 + - Ubuntu 20.04 LTS + - dpdk 19.11.7-0ubuntu0.20.04.1 + (we tested it with 19.11.10~rc1, but the problem persists) + - Intel XXV710 + - Cisco 25G AOC cables - We are seeing issues with link status of ports as DPDK-bond members and those links suddenly go away and marked as down. There are multiple parameters that could cause this issue, but one of the suggestions we've got from a server vendor was that the following upstream patch would be required to support 25G AOC/ACC cables. - https://github.com/DPDK/dpdk/commit/b1daa34614 - (available for v21.05 onward) + Patch to backport: + https://git.dpdk.org/dpdk/commit/?id=b1daa3461429e7674206a714c17adca65e9b44b4 - (todo: define a test case) + [Impact] - [expected status] - ---- dpdk-bond0 ---- - bond_mode: balance-tcp - bond may use recirculation: yes, Recirc-ID : 1 - bond-hash-basis: 0 - updelay: 0 ms - downdelay: 0 ms - next rebalance: 1691 ms - lacp_status: negotiated - lacp_fallback_ab: false - active slave mac: 40:a6:b7:3e:4a:60(dpdk-d2cb784) - slave dpdk-7272e20: enabled - may_enable: true - slave dpdk-d2cb784: enabled - active slave - may_enable: true + DPDK ports for a bond get disabled and no traffic goes in and out after + openvswitch restart with the combination above. If that happens the DPDK + bond has to be re-created as a workaround but it's not feasible since + service restart basically breaks everything. - - [after sometime - links are lost] ---- dpdk-bond0 ---- bond_mode: balance-tcp bond may use recirculation: yes, Recirc-ID : 1 bond-hash-basis: 0 updelay: 0 ms downdelay: 0 ms next rebalance: 7267 ms lacp_status: configured lacp_fallback_ab: false active slave mac: 00:00:00:00:00:00(none) slave dpdk-7272e20: disabled may_enable: false slave dpdk-d2cb784: disabled may_enable: false + + [Test Plan] + + 1. configure a DPDK bond with openvswitch as follows for example. + + $ sudo ovs-appctl bond/show dpdk-bond0 + + ---- dpdk-bond0 ---- + bond_mode: balance-tcp + bond may use recirculation: yes, Recirc-ID : 1 + bond-hash-basis: 0 + updelay: 0 ms + downdelay: 0 ms + next rebalance: 1691 ms + lacp_status: negotiated + lacp_fallback_ab: false + active slave mac: 40:a6:b7:XX:YY:ZZ(dpdk-d2cb784) + slave dpdk-7272e20: enabled + may_enable: true + slave dpdk-d2cb784: enabled + active slave + may_enable: true + + 2. Apply updated packages + + 3. Reboot the machine (just to make sure we are not using anything old) + + 4. Restart the openvswitch + + $ sudo systemctl restart openvswitch-switch + + 5. Confirm ports are enabled after both the step 3. and 4. and the port + status matches the one in the step 1. + + + [Where problems could occur] + + The scope of the patch is i40e and the two specific cable types only: + i40e + 25G AOC and ACC cables so it's unlikely to affect any other + combinations. Before this patch, 25G AOC/ACC cables were not in the + additional PHY types of the driver functionality so it's not likely to + make things worse.
-- You received this bug notification because you are a member of Ubuntu Bugs, which is subscribed to Ubuntu. https://bugs.launchpad.net/bugs/1940957 Title: DPDK ports get disabled after Open vSwitch restart with Intel XXV710(i40e) and 25G AOC cables To manage notifications about this bug go to: https://bugs.launchpad.net/ubuntu/+source/dpdk/+bug/1940957/+subscriptions -- ubuntu-bugs mailing list ubuntu-bugs@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs