On Wed, Jun 22, 2016 at 12:02 PM, fabrice grelaud <fabrice.grel...@u-bordeaux.fr> wrote: > > Le 22 juin 2016 à 17:35, fabrice grelaud <fabrice.grel...@u-bordeaux.fr> a > écrit : > > > Le 22 juin 2016 à 15:45, Assaf Muller <as...@redhat.com> a écrit : > > On Wed, Jun 22, 2016 at 9:24 AM, fabrice grelaud > <fabrice.grel...@u-bordeaux.fr> wrote: > > Hi, > > we deployed our openstack infrastructure with your « exciting » project > openstack-ansible (mitaka 13.1.2) but we have some problems with L3HA after > create router. > > Our infra (closer to the doc): > 3 controllers nodes (with bond0 (br-mgmt, br-storage), bond1 (br-vxlan, > br-vlan)) > 2 compute nodes (same for network) > > We create an external network (vlan type), an internal network (vxlan type) > and a router connected to both networks. > And when we launch an instance (cirros), we can’t receive an ip on the vm. > > We have: > > root@p-osinfra03-utility-container-783041da:~# neutron > l3-agent-list-hosting-router router-bim > +--------------------------------------+-----------------------------------------------+----------------+-------+----------+ > | id | host > | admin_state_up | alive | ha_state | > +--------------------------------------+-----------------------------------------------+----------------+-------+----------+ > | 3c7918e5-3ad6-4f82-a81b-700790e3c016 | > p-osinfra01-neutron-agents-container-f1ab9c14 | True | :-) | > active | > | f2bf385a-f210-4dbc-8d7d-4b7b845c09b0 | > p-osinfra02-neutron-agents-container-48142ffe | True | :-) | > active | > | 55350fac-16aa-488e-91fd-a7db38179c62 | > p-osinfra03-neutron-agents-container-2f6557f0 | True | :-) | > active | > +--------------------------------------+-----------------------------------------------+----------------+-------+—————+ > > I know, i got a problem now because i should have :-) active, :-) standby, > :-) standby… Snif... > > root@p-osinfra01-neutron-agents-container-f1ab9c14:~# ip netns > qrouter-eeb2147a-5cc6-4b5e-b97c-07cfc141e8e6 > qdhcp-0ba266fb-15c4-4566-ae88-92d4c8fd2036 > > root@p-osinfra01-neutron-agents-container-f1ab9c14:~# ip netns exec > qrouter-eeb2147a-5cc6-4b5e-b97c-07cfc141e8e6 ip a sh > 1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN group > default > link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00 > inet 127.0.0.1/8 scope host lo > valid_lft forever preferred_lft forever > inet6 ::1/128 scope host > valid_lft forever preferred_lft forever > 2: ha-4a5f0287-91@if6: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1450 qdisc > pfifo_fast state UP group default qlen 1000 > link/ether fa:16:3e:c2:67:a9 brd ff:ff:ff:ff:ff:ff > inet 169.254.192.1/18 brd 169.254.255.255 scope global ha-4a5f0287-91 > valid_lft forever preferred_lft forever > inet 169.254.0.1/24 scope global ha-4a5f0287-91 > valid_lft forever preferred_lft forever > inet6 fe80::f816:3eff:fec2:67a9/64 scope link > valid_lft forever preferred_lft forever > 3: qr-44804d69-88@if9: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1450 qdisc > pfifo_fast state UP group default qlen 1000 > link/ether fa:16:3e:a5:8c:f2 brd ff:ff:ff:ff:ff:ff > inet 192.168.100.254/24 scope global qr-44804d69-88 > valid_lft forever preferred_lft forever > inet6 fe80::f816:3eff:fea5:8cf2/64 scope link > valid_lft forever preferred_lft forever > 4: qg-c5c7378e-1d@if12: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc > pfifo_fast state UP group default qlen 1000 > link/ether fa:16:3e:b6:4c:97 brd ff:ff:ff:ff:ff:ff > inet 147.210.240.11/23 scope global qg-c5c7378e-1d > valid_lft forever preferred_lft forever > inet 147.210.240.12/32 scope global qg-c5c7378e-1d > valid_lft forever preferred_lft forever > inet6 fe80::f816:3eff:feb6:4c97/64 scope link > valid_lft forever preferred_lft forever > > Same result on infra02 and infra03, qr and qg interfaces have the same ip, > and ha interfaces the address 169.254.0.1. > > If we stop 2 neutron agent containers (p-osinfra02, p-osinfra03) and we > restart the first (p-osinfra01), we can reboot the instance and we got an > ip, a floating ip and we can access by ssh from internet to the vm. (Note: > after few time, we loss our connectivity too). > > But if we restart the two containers, we got a ha_state to « standby » until > the three become « active » and finally we have the problem again. > > The three routers on infra 01/02/03 are seen as master. > > If we ping from our instance to the router (internal network 192.168.100.4 > to 192.168.100.254) we can see some ARP Request > ARP, Request who-has 192.168.100.254 tell 192.168.100.4, length 28 > ARP, Request who-has 192.168.100.254 tell 192.168.100.4, length 28 > ARP, Request who-has 192.168.100.254 tell 192.168.100.4, length 28 > > And on the compute node we see all these frames on the various interfaces > tap / vxlan-89 / br-vxlan / bond1.vxlanvlan / bond1 / em2 but nothing back. > > We also have on ha interface, on each router, the VRRP communication > (heartbeat packets over a hidden project network that connects all ha > routers (vxlan 70) ) . Priori as normal, each router thinks to be master. > > root@p-osinfra01-neutron-agents-container-f1ab9c14:~# ip netns exec > qrouter-eeb2147a-5cc6-4b5e-b97c-07cfc141e8e6 tcpdump -nl -i ha-4a5f0287-91 > tcpdump: verbose output suppressed, use -v or -vv for full protocol decode > listening on ha-4a5f0287-91, link-type EN10MB (Ethernet), capture size 65535 > bytes > IP 169.254.192.1 > 224.0.0.18: VRRPv2, Advertisement, vrid 1, prio 50, > authtype simple, intvl 2s, length 20 > IP 169.254.192.1 > 224.0.0.18: VRRPv2, Advertisement, vrid 1, prio 50, > authtype simple, intvl 2s, length 20 > IP 169.254.192.1 > 224.0.0.18: VRRPv2, Advertisement, vrid 1, prio 50, > authtype simple, intvl 2s, length 20 > IP 169.254.192.1 > 224.0.0.18: VRRPv2, Advertisement, vrid 1, prio 50, > authtype simple, intvl 2s, length 20 > > root@p-osinfra02-neutron-agents-container-48142ffe:~# ip netns exec > qrouter-eeb2147a-5cc6-4b5e-b97c-07cfc141e8e6 tcpdump -nt -i ha-4ee5f8d0-7f > tcpdump: verbose output suppressed, use -v or -vv for full protocol decode > listening on ha-4ee5f8d0-7f, link-type EN10MB (Ethernet), capture size 65535 > bytes > IP 169.254.192.3 > 224.0.0.18: VRRPv2, Advertisement, vrid 1, prio 50, > authtype simple, intvl 2s, length 20 > IP 169.254.192.3 > 224.0.0.18: VRRPv2, Advertisement, vrid 1, prio 50, > authtype simple, intvl 2s, length 20 > IP 169.254.192.3 > 224.0.0.18: VRRPv2, Advertisement, vrid 1, prio 50, > authtype simple, intvl 2s, length 20 > IP 169.254.192.3 > 224.0.0.18: VRRPv2, Advertisement, vrid 1, prio 50, > authtype simple, intvl 2s, length 20 > IP 169.254.192.3 > 224.0.0.18: VRRPv2, Advertisement, vrid 1, prio 50, > authtype simple, intvl 2s, length 20 > > > Are you seeing VRRP advertisements crossing nodes though? That tcpdump > only shows advertisements from the local node. If nodes aren't > receiving VRRP messages from other nodes, keepalived will declare > itself as master (As expected). Can you ping the 'ha' interface from > one router namespace to the other? > > > I stop the three neutron agent container. > Restart on infra01 then on infra02 > > I can see VRRP frames from infra01 (169.254.192.1 -> 224.0.0.18) and receive > by infra02. > > root@p-osinfra02:~# tcpdump -nl -i em2 | grep 169.254 > tcpdump: WARNING: em2: no IPv4 address assigned > tcpdump: verbose output suppressed, use -v or -vv for full protocol decode > listening on em2, link-type EN10MB (Ethernet), capture size 65535 bytes > IP 169.254.192.1 > 224.0.0.18: VRRPv2, Advertisement, vrid 1, prio 50, > authtype simple, intvl 2s, length 20 > IP 169.254.192.1 > 224.0.0.18: VRRPv2, Advertisement, vrid 1, prio 50, > authtype simple, intvl 2s, length 20 > …. > …. > then i have > IP 169.254.192.3 > 224.0.0.18: VRRPv2, Advertisement, vrid 1, prio 50, > authtype simple, intvl 2s, length 20 > > No more 169.254.192.1 from infra01 but the IP of HA interface from router on > infra02. > > And no more VRRP advertisements cross the nodes. > On each infra node, we see VRRP advertisements from the node itself but > nothing from the other. > > And otherwise, i can ping ha interface from one router namespace to the > other: > root@p-osinfra01-neutron-agents-container-f1ab9c14:~# ip netns exec > qrouter-eeb2147a-5cc6-4b5e-b97c-07cfc141e8e6 ping 169.254.192.3 > PING 169.254.192.3 (169.254.192.3) 56(84) bytes of data. > 64 bytes from 169.254.192.3: icmp_seq=1 ttl=64 time=0.297 ms > 64 bytes from 169.254.192.3: icmp_seq=2 ttl=64 time=0.239 ms > 64 bytes from 169.254.192.3: icmp_seq=3 ttl=64 time=0.264 ms > > im’ going to test with other version of keepalived (current version here > 1.2.7-1 ubuntu 14.04). > > Thanks to help > > > Note: > I said i can ping between ha interface but not for long time. At one point, > i can’t anymore… :-(
That's the problem. This becomes a normal Neutron troubleshooting: Why can't one port ping the other? This might help: https://assafmuller.com/2015/08/31/neutron-troubleshooting/ > > > > > > Someone could tell me if he has already encountered this problem ? > The infra and compute nodes are connected to a nexus 9000 switch. > > Thank you in advance for taking the time to study my request. > > Fabrice Grelaud > Université de Bordeaux > > > __________________________________________________________________________ > OpenStack Development Mailing List (not for usage questions) > Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe > http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev > > > __________________________________________________________________________ > OpenStack Development Mailing List (not for usage questions) > Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe > http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev > > > __________________________________________________________________________ > OpenStack Development Mailing List (not for usage questions) > Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe > http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev > > > > __________________________________________________________________________ > OpenStack Development Mailing List (not for usage questions) > Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe > http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev > __________________________________________________________________________ OpenStack Development Mailing List (not for usage questions) Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
