Re: VRRP entering master state on both nodes in pair

Answers are in-line below.

On 4/22/15 1:38 AM, Tore Anderson wrote:
> * "Todd D. Taft" <[email protected]>
>
>> Something isn't working as I would expect with my configuration, because
>> shortly after starting keepalived on either system, that system thinks
>> it's the master for both services (even when the other system is up).
>> As such, all requests for both services go to whichever system sent out
>> the arp packets most recently.
> Hi Todd,
>
> Check if they two hosts are seeing each others' VRRP advertisements,
> for example with with "tcpdump -vni eth0 vrrp". If they don't, check if
> your switch has multicast snooping enabled and if so try to disable it.
I'm seeing packets from both systems:
Running tcpdump on hyde (primary IP 10.1.42.243):
     10.1.42.242 > 224.0.0.18: vrrp 10.1.42.242 > 224.0.0.18: VRRPv2, 
Advertisement, vrid 41, prio 105, authtype none, intvl 1s, length 20, 
addrs: 10.1.42.240
20:21:34.403274 IP (tos 0xc0, ttl 255, id 145, offset 0, flags [none], 
proto VRRP (112), length 40)
     10.1.42.243 > 224.0.0.18: vrrp 10.1.42.243 > 224.0.0.18: VRRPv2, 
Advertisement, vrid 41, prio 106, authtype none, intvl 1s, length 20, 
addrs: 10.1.42.240
20:21:34.403782 IP (tos 0xc0, ttl 255, id 145, offset 0, flags [none], 
proto VRRP (112), length 40)
     10.1.42.243 > 224.0.0.18: vrrp 10.1.42.243 > 224.0.0.18: VRRPv2, 
Advertisement, vrid 42, prio 105, authtype none, intvl 1s, length 20, 
addrs: 10.1.42.241
20:21:34.800226 IP (tos 0xc0, ttl 255, id 140, offset 0, flags [none], 
proto VRRP (112), length 40)
     10.1.42.242 > 224.0.0.18: vrrp 10.1.42.242 > 224.0.0.18: VRRPv2, 
Advertisement, vrid 42, prio 106, authtype none, intvl 1s, length 20, 
addrs: 10.1.42.241
20:21:34.801374 IP (tos 0xc0, ttl 255, id 140, offset 0, flags [none], 
proto VRRP (112), length 40)

(Jekyll's primary IP Is 10.1.42.242).  (I see similar results running 
the tcpdump on jekyll.)
> Do eth0 on both hosts have an IPv4 address/prefix configured that
> encompasses 10.1.42.240 and .241? If not, try disabling uRPF with
> "sysctl -w net/ipv4/conf/eth0/rp_filter=0".
Both systems are in the same VLAN and within IP broadcast range of each 
other  (netmask 255.255.254.0).
> Why do the instances in state MASTER have a lower priority than the
> corresponding instance on the other node with state BACKUP? That seems
> backwards to me. For what it's worth, I prefer to let all instances
> start out with state BACKUP (they'll elect a master soon enough).
I've switched all of the MASTER entries to BACKUP in the config file, 
but I'm still seeing both systems quickly moving into MASTER status for 
both addresses shortly after restarting keepalived.
>
> What is the purpose of your track_scripts? They seem to me to only ping
> the local host, which should always succeed as long as the host is up,
> right?
It's sort of a placeholder.  Ideally, it would be replaced with 
something that actually checks that the service is running/responding on 
the remote system (e.g. a nagios-style check script).  I was trying to 
get things working with a simple ping before adding a more complex 
service check.
>
> Tore
>
Thanks for your help,
--Todd
> -- 
> Todd D. Taft
> [email protected]

------------------------------------------------------------------------------
BPM Camp - Free Virtual Workshop May 6th at 10am PDT/1PM EDT
Develop your own process in accordance with the BPMN 2 standard
Learn Process modeling best practices with Bonita BPM through live exercises
http://www.bonitasoft.com/be-part-of-it/events/bpm-camp-virtual- event?utm_
source=Sourceforge_BPM_Camp_5_6_15&utm_medium=email&utm_campaign=VA_SF