[openstack-dev] [openstack-ansible] L3HA problem
Assaf Muller
assaf at redhat.com
Wed Jun 22 17:40:51 UTC 2016
On Wed, Jun 22, 2016 at 12:02 PM, fabrice grelaud
<fabrice.grelaud at u-bordeaux.fr> wrote:
>
> Le 22 juin 2016 à 17:35, fabrice grelaud <fabrice.grelaud at u-bordeaux.fr> a
> écrit :
>
>
> Le 22 juin 2016 à 15:45, Assaf Muller <assaf at redhat.com> a écrit :
>
> On Wed, Jun 22, 2016 at 9:24 AM, fabrice grelaud
> <fabrice.grelaud at 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 at 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 at p-osinfra01-neutron-agents-container-f1ab9c14:~# ip netns
> qrouter-eeb2147a-5cc6-4b5e-b97c-07cfc141e8e6
> qdhcp-0ba266fb-15c4-4566-ae88-92d4c8fd2036
>
> root at 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 at 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 at 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 at 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 at 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 at 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 at 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 at 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-request at 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-request at 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-request at 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-request at lists.openstack.org?subject:unsubscribe
> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
>
More information about the OpenStack-dev
mailing list