[openstack-dev] [openstack-ansible] L3HA problem
fabrice grelaud
fabrice.grelaud at u-bordeaux.fr
Fri Jun 24 08:55:36 UTC 2016
> Le 22 juin 2016 à 19:40, Assaf Muller <assaf at redhat.com> a écrit :
>
> On Wed, Jun 22, 2016 at 12:02 PM, fabrice grelaud
> <fabrice.grelaud at u-bordeaux.fr <mailto: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/ <https://assafmuller.com/2015/08/31/neutron-troubleshooting/>
Hi,
thanks for the link. I already had a look (more or less) and finally, i suspected a problem rather on switch side. (nexus)
And after some investigation and tcpdump, we saw that packets to 239.1.1.1 are not forwarding by the switch.
In fact, « igmp snooping » is enabled by default on nexus switch causing the bad behaviour.
We disable igmp snooping and That’s all folks !
Cordially,
>
>>
>>
>>
>>
>>
>> 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
>>
>
> __________________________________________________________________________
> OpenStack Development Mailing List (not for usage questions)
> Unsubscribe: OpenStack-dev-request at lists.openstack.org <mailto:OpenStack-dev-request at lists.openstack.org>?subject:unsubscribe
> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev <http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack-dev/attachments/20160624/6d34b70e/attachment.html>
More information about the OpenStack-dev
mailing list