CC13: Restarting of vRouter Agent (DPDK) causes the vif sockets to be removed from /var/run/vrouter

[Steven Sciriha]

Issue Description:

During our deployment testing we were required to do an Overcloud heat stack update. Given the nature of the configuration change this required all DPDK vRouter agents to be restarted (soft reset).

Soon after that we quickly noticed that there were connectivity issues with all the DPDK VMs running on the vRouter DPDK computes.

Investigation revealed that while the DPDK vRouter was still showing the VM tap interfaces when running the "vif --list" the associated userspace sockets were not listed under /var/run/vrouter.

I was not able to reproduce the same issue on kernel-mode compute vRouter.

Steps to reproduce:

1. On DPDK compute confirm tap interface is displayed using vif command

# vif --list

vif0/0 PCI: 0000:00:00.0 (Speed 20000, Duplex 0)
            Type:Physical HWaddr:68:05:ca:42:6b:58 IPaddr:0.0.0.0
            Vrf:0 Mcast Vrf:65535 Flags:TcL3L2VpEr QOS:-1 Ref:180
            RX device packets:460731944 bytes:227325677160 errors:0
            RX port packets:468848 errors:0
            RX queue errors to lcore 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
            RX packets:468193 bytes:94599396 errors:0
            TX packets:63124 bytes:7204203 errors:0
            Drops:43826
            TX port packets:63106 errors:18
            TX device packets:400643180 bytes:71087343258 errors:0

vif0/1 PMD: vhost0
            Type:Host HWaddr:68:05:ca:42:6b:58 IPaddr:100.64.144.34
            Vrf:0 Mcast Vrf:65535 Flags:L3DEr QOS:-1 Ref:17
            RX device packets:18670 bytes:1506345 errors:0
            RX queue packets:18670 errors:0
            RX queue errors to lcore 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
            RX packets:18670 bytes:1506345 errors:0
            TX packets:392862 bytes:82763018 errors:0
            Drops:0
            TX queue packets:55373 errors:337489
            TX device packets:55373 bytes:19531591 errors:0

vif0/2 Socket: unix
            Type:Agent HWaddr:00:00:5e:00:01:00 IPaddr:0.0.0.0
            Vrf:65535 Mcast Vrf:65535 Flags:L3Er QOS:-1 Ref:3
            RX port packets:44656 errors:0
            RX queue errors to lcore 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
            RX packets:44656 bytes:7670834 errors:0
            TX packets:75703 bytes:12401925 errors:0
            Drops:0

vif0/3 PMD: tap27b5ab41-e5
            Type:Virtual HWaddr:00:00:5e:00:01:00 IPaddr:172.22.65.4
            Vrf:2 Mcast Vrf:2 Flags:PL3L2DEr QOS:-1 Ref:20
            RX port packets:331 errors:0 syscalls:1
            RX queue packets:53 errors:0
            RX queue errors to lcore 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
            RX packets:331 bytes:16716 errors:0
            TX packets:376 bytes:19138 errors:0
            ISID: 0 Bmac: 02:27:b5:ab:41:e5
            Drops:52
            TX port packets:360 errors:16 syscalls:353

2. Ensure associated socket is present under /var/run/vrouter/

# ls /var/run/vrouter/ | grep tap
uvh_vif_tap27b5ab41-e5

3. Restart contrail vrouter agent container

# docker restart contrail_vrouter_agent
contrail_vrouter_agent

4. Confirm the tap socket is no longer plumbed:

# ls /var/run/vrouter/ | grep tap
#

Workaround:

Restart impacted VM/nova instance using "openstack server stop/start"

1. Confirm the tap socket is plumbed again after VM restart

# ls /var/run/vrouter/ | grep tap
uvh_vif_tap27b5ab41-e5