Activity log for bug #1582452

Date Who What changed Old value New value Message
2016-05-17 00:06:41 vivekananda shenoy bug added bug
2016-05-17 00:07:24 vivekananda shenoy description Something we saw during VES troubleshooting call today. In 3.0 release whenever a Control node receives multiple BGP path for the same IP prefix from another Contrail controller or from the MX the bgp best path check is done by comparing local-preference and in case if one of the path has shorter AS-PATH length compared to the other one both are treated equal. VES topo is as shown below. Cluster1 is getting 3 paths for IP address of a VM in cluster 2 and all 3 paths are installed in the agent routing table although the direct EBGP path from cluster 2 has AS-PATH-LEN of 1. This is creating issues w.r.t ARP resolution for the IP address of VM in cluster 2, as the ARP is getting resolved to vhost0 MAC address of the compute node in cluster 2. I am opening a separate PR for that. Regards, Vivek +------+ | | +----------+ MX1 +-------------+ | | | | | +------+ | | +------+ | | | | | | | MX2 | | EBGP | +------+ +--------+ | | | +------+ | | | | | | | | | | | |EBGP EBGP | |EBGP | | | | | | | | +-----+---+-+ ++----+------+ | | | | | | EBGP | | | Cluster1+-------------------+ Cluster2 | | | | | +-----------+ +------------+ Something we saw during VES troubleshooting call today. In 3.0 release whenever a Control node receives multiple BGP path for the same IP prefix from another Contrail controller or from the MX the bgp best path check is done by comparing local-preference and in case if one of the path has shorter AS-PATH length compared to the other one both are treated equal. VES topo is attached. Cluster1 is getting 3 paths for IP address of a VM in cluster 2 and all 3 paths are installed in the agent routing table although the direct EBGP path from cluster 2 has AS-PATH-LEN of 1. This is creating issues w.r.t ARP resolution for the IP address of VM in cluster 2, as the ARP is getting resolved to vhost0 MAC address of the compute node in cluster 2. I am opening a separate PR for that. Regards, Vivek
2016-05-17 00:08:56 vivekananda shenoy attachment added VES-topo1.txt https://bugs.launchpad.net/juniperopenstack/+bug/1582452/+attachment/4664163/+files/VES-topo1.txt
2016-05-17 16:34:53 Nischal Sheth description Something we saw during VES troubleshooting call today. In 3.0 release whenever a Control node receives multiple BGP path for the same IP prefix from another Contrail controller or from the MX the bgp best path check is done by comparing local-preference and in case if one of the path has shorter AS-PATH length compared to the other one both are treated equal. VES topo is attached. Cluster1 is getting 3 paths for IP address of a VM in cluster 2 and all 3 paths are installed in the agent routing table although the direct EBGP path from cluster 2 has AS-PATH-LEN of 1. This is creating issues w.r.t ARP resolution for the IP address of VM in cluster 2, as the ARP is getting resolved to vhost0 MAC address of the compute node in cluster 2. I am opening a separate PR for that. Regards, Vivek In 3.0 release whenever a Control node receives multiple BGP path for the same IP prefix from another Contrail controller or from the MX the bgp best path check is done by comparing local-preference and in case if one of the path has shorter AS-PATH length compared to the other one both are treated equal. VES topo is attached. Cluster1 is getting 3 paths for IP address of a VM in cluster 2 and all 3 paths are installed in the agent routing table although the direct EBGP path from cluster 2 has AS-PATH-LEN of 1. This is creating issues w.r.t ARP resolution for the IP address of VM in cluster 2, as the ARP is getting resolved to vhost0 MAC address of the compute node in cluster 2. I am opening a separate PR for that.
2016-05-17 16:35:05 Nischal Sheth juniperopenstack: importance Undecided High
2016-05-17 16:35:07 Nischal Sheth juniperopenstack: assignee Nischal Sheth (nsheth)
2016-05-17 16:36:38 Nischal Sheth nominated for series juniperopenstack/r3.0
2016-05-17 16:36:38 Nischal Sheth bug task added juniperopenstack/r3.0
2016-05-17 16:36:38 Nischal Sheth nominated for series juniperopenstack/trunk
2016-05-17 16:36:38 Nischal Sheth bug task added juniperopenstack/trunk
2016-05-17 16:36:45 Nischal Sheth juniperopenstack/r3.0: assignee Nischal Sheth (nsheth)
2016-05-17 16:36:47 Nischal Sheth juniperopenstack/r3.0: importance Undecided High
2016-05-17 17:28:30 Nischal Sheth summary 3.0 Controller : There has to be check against the AS path length when comparing ECMP routes in the case of federated clusters ECMP logic should compare AS path length in case of federated clusters
2016-05-17 17:29:07 Nischal Sheth summary ECMP logic should compare AS path length in case of federated clusters ECMP logic should compare AS path length
2016-05-17 17:30:03 Nischal Sheth description In 3.0 release whenever a Control node receives multiple BGP path for the same IP prefix from another Contrail controller or from the MX the bgp best path check is done by comparing local-preference and in case if one of the path has shorter AS-PATH length compared to the other one both are treated equal. VES topo is attached. Cluster1 is getting 3 paths for IP address of a VM in cluster 2 and all 3 paths are installed in the agent routing table although the direct EBGP path from cluster 2 has AS-PATH-LEN of 1. This is creating issues w.r.t ARP resolution for the IP address of VM in cluster 2, as the ARP is getting resolved to vhost0 MAC address of the compute node in cluster 2. I am opening a separate PR for that. Version 3.0.2.0 build 25 If CN receives multiple BGP path for the same IP prefix from another Contrail controller or from the MX the bgp best path check is done by comparing local-preference and in case if one of the path has shorter AS-PATH length compared to the other one both are treated equal. VES topo is attached. Cluster1 is getting 3 paths for IP address of a VM in cluster 2 and all 3 paths are installed in the agent routing table although the direct EBGP path from cluster 2 has AS-PATH-LEN of 1. This is creating issues w.r.t ARP resolution for the IP address of VM in cluster 2, as the ARP is getting resolved to vhost0 MAC address of the compute node in cluster 2. I am opening a separate PR for that.
2016-05-17 17:30:23 Nischal Sheth tags federation ves contrail-control federation ves
2016-05-18 22:38:18 vivekananda shenoy tags contrail-control federation ves blocker contrail-control federation ves
2016-05-20 06:10:55 Nischal Sheth information type Proprietary Public
2016-05-20 18:46:24 Jeba Paulaiyan juniperopenstack/r3.0: milestone r3.0.2.0
2016-05-24 04:45:14 Ashish Ranjan juniperopenstack/r3.0: status New Won't Fix
2016-05-24 16:50:06 vivekananda shenoy tags blocker contrail-control federation ves contrail-control federation ves
2016-05-24 21:39:16 OpenContrail Admin juniperopenstack/trunk: status New In Progress
2016-05-24 21:59:32 Nischal Sheth juniperopenstack/r3.0: status Won't Fix New
2016-05-24 22:03:29 OpenContrail Admin juniperopenstack/r3.0: status New In Progress
2016-05-25 07:02:13 OpenContrail Admin juniperopenstack/r3.0: status In Progress Fix Committed
2016-05-29 05:27:38 OpenContrail Admin juniperopenstack/trunk: status In Progress Fix Committed
2016-05-29 05:27:39 OpenContrail Admin juniperopenstack/trunk: milestone r3.1.0.0-fcs