VIP interfaces in the same IP subnet send duplicate ARP replies

Bug #1594722 reported by Simon Pasquier on 2016-06-21
6
This bug affects 1 person
Affects Status Importance Assigned to Milestone
StackLight
Undecided
Swann Croiset

Bug Description

Tested on MOS 9 and the plugins from the master branch.

The InfluxDB and Elasticsearch are located on the same node and share the same subnet.

# ip netns exec haproxy ip r
default via 240.0.0.1 dev hapr-ns metric 10000
10.109.10.0/24 via 240.0.0.1 dev hapr-ns metric 10000
10.109.11.0/24 dev b_influxdb proto kernel scope link src 10.109.11.13
10.109.11.0/24 dev b_es_vip_mgmt proto kernel scope link src 10.109.11.15
10.109.11.0/24 via 240.0.0.1 dev hapr-ns metric 10000
10.109.12.0/24 via 240.0.0.1 dev hapr-ns metric 10000
240.0.0.0/30 dev hapr-ns proto kernel scope link src 240.0.0.2

# ip netns exec haproxy ip l
1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN mode DEFAULT group default
    link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
14: b_influxdb: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc pfifo_fast state UP mode DEFAULT group default qlen 1000
    link/ether f2:6c:c9:99:c9:b6 brd ff:ff:ff:ff:ff:ff
16: hapr-ns: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc pfifo_fast state UP mode DEFAULT group default qlen 1000
    link/ether ca:6d:4c:bf:42:25 brd ff:ff:ff:ff:ff:ff
18: b_es_vip_mgmt: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc pfifo_fast state UP mode DEFAULT group default qlen 1000
    link/ether a6:d5:de:15:e4:7c brd ff:ff:ff:ff:ff:ff

From another node, I can see that both interfaces reply to ARP requests for the Elasticsearch VIP.

# arping -b -I br-mgmt 10.109.11.15
ARPING 10.109.11.15 from 10.109.11.5 br-mgmt
Unicast reply from 10.109.11.15 [A6:D5:DE:15:E4:7C] 1.213ms
Unicast reply from 10.109.11.15 [F2:6C:C9:99:C9:B6] 1.256ms
Unicast reply from 10.109.11.15 [A6:D5:DE:15:E4:7C] 1.291ms
Unicast reply from 10.109.11.15 [F2:6C:C9:99:C9:B6] 1.338ms
Unicast reply from 10.109.11.15 [A6:D5:DE:15:E4:7C] 0.897ms
Unicast reply from 10.109.11.15 [F2:6C:C9:99:C9:B6] 1.002ms

root@node-1:~# arping -I br-mgmt 10.109.11.15
ARPING 10.109.11.15 from 10.109.11.5 br-mgmt
Unicast reply from 10.109.11.15 [A6:D5:DE:15:E4:7C] 1.626ms
Unicast reply from 10.109.11.15 [F2:6C:C9:99:C9:B6] 1.647ms
Unicast reply from 10.109.11.15 [F2:6C:C9:99:C9:B6] 1.582ms

root@node-1:~# arping -I br-mgmt 10.109.11.13
ARPING 10.109.11.13 from 10.109.11.5 br-mgmt
Unicast reply from 10.109.11.13 [A6:D5:DE:15:E4:7C] 1.382ms
Unicast reply from 10.109.11.13 [F2:6C:C9:99:C9:B6] 1.418ms
Unicast reply from 10.109.11.13 [F2:6C:C9:99:C9:B6] 0.884ms

Changed in lma-toolchain:
assignee: nobody → Swann Croiset (swann-w)
status: New → In Progress

Reviewed: https://review.openstack.org/331737
Committed: https://git.openstack.org/cgit/openstack/fuel-plugin-lma-infrastructure-alerting/commit/?id=a70862ce2bf113949597e6ec4a4912c34e8aa251
Submitter: Jenkins
Branch: master

commit a70862ce2bf113949597e6ec4a4912c34e8aa251
Author: Swann Croiset <email address hidden>
Date: Mon Jun 20 17:10:05 2016 +0200

    Support multiple Virtual IPs on the same subnet

    Fixes-bug: #1594722

    Change-Id: I825009dbc2cc3a656c6dbd1b89cd34d77bb7ede3

Changed in lma-toolchain:
status: In Progress → Fix Committed

Reviewed: https://review.openstack.org/345269
Committed: https://git.openstack.org/cgit/openstack/fuel-plugin-lma-infrastructure-alerting/commit/?id=de3cc5e66375546272ea30db41d44fc49aa50c72
Submitter: Jenkins
Branch: master

commit de3cc5e66375546272ea30db41d44fc49aa50c72
Author: Simon Pasquier <email address hidden>
Date: Thu Jul 21 11:03:54 2016 +0200

    Add ordering constraints between Apache and VIPs

    This ensures that the Nagios Apache service won't start until the VIP
    interfaces are up and running. This is required because the Apache OCF
    script needs to tweak the ARP parameters of the VIP interfaces.

    Change-Id: I4cc7b7db06cda31064b22e0283af478228202b54
    Related-Bug: #1594722

Reviewed: https://review.openstack.org/345897
Committed: https://git.openstack.org/cgit/openstack/fuel-plugin-lma-infrastructure-alerting/commit/?id=584d2f0eea96c773741f752cbaaeaecb71c00bed
Submitter: Jenkins
Branch: stable/0.10

commit 584d2f0eea96c773741f752cbaaeaecb71c00bed
Author: Simon Pasquier <email address hidden>
Date: Thu Jul 21 11:03:54 2016 +0200

    Add ordering constraints between Apache and VIPs

    This ensures that the Nagios Apache service won't start until the VIP
    interfaces are up and running. This is required because the Apache OCF
    script needs to tweak the ARP parameters of the VIP interfaces.

    Change-Id: I4cc7b7db06cda31064b22e0283af478228202b54
    Related-Bug: #1594722

lustack (lustack) wrote :

This issue still exist and the similar strategy will be used in haproxy netns.

https://review.openstack.org/#/c/496475/

Please review this change list.

lustack (lustack) on 2017-08-31
Changed in lma-toolchain:
status: Fix Committed → In Progress
To post a comment you must log in.
This report contains Public information  Edit
Everyone can see this information.

Other bug subscribers