Fatcharly,
I have seen a lot of approaches of HA solution for pound. The best one
in my point of view, which I have used in the last 3 years, have caused me
almost no problem and is a very effective one. You simply use a crude
heartbeat. What I mean crude heartbeat? Do not use corosync, pacemaker,etc....
Only the heartbeat (version 3 is a good choice). You will operate with
heartbeat not with an advanced CRM (cluster resource manager) like pacemaker,
but with its very simple CRM which is enough for a good HA solution.You will be
limited to 2 nodes , what is not a problem in my point of view.
You do not need to touch the sysctl.conf file. Your will choose a
virtual ip for the cluster and put it in the heartbeat files(not in the network
config files).Just like my /etc/ha.d/haresources file (just one line, machine
www12a, vip 172.31.3.97):
www12a 172.31.3.97
My ha.cf file:
logfile /var/log/ha-log
logfacility local0
keepalive 3
deadtime 15
warntime 10
initdead 60
udpport 694
#bcast eth2 eth0
ucast eth0 172.31.3.202
ucast eth1 192.168.52.200
ucast eth2 10.0.0.1
ucast eth0 172.31.3.203
ucast eth1 192.168.52.201
ucast eth2 10.0.0.2
auto_failback off
node www12a
node www12b
After that you will keep Pound ACTIVE on the 2 machines with the SAME config
files.
DO not forget to use more than one communication path between the two nodes, as
you can see in the ha.cf file example above.
A cross cable is a perfect choice as an additional communication path.
The need of a fencing device will not be critical as you will put redundant
communication paths.
The init configuration of the system(in both machines) must start pound and
heartbeat.
Avoid a firewall between the two nodes.
Regards,
Roberto.
-----Mensagem original-----
De: Joe Gooch [mailto:[email protected]]
Enviada em: quinta-feira, 19 de abril de 2012 10:51
Para: '[email protected]'
Assunto: RE: [Pound Mailing List] pound with heartbeat/pacemaker
I use option 1.
In your sysctl.conf, add:
#Pound stuff binds even if we aren't primary
net.ipv4.ip_nonlocal_bind=1
Activate it immediately with
sysctl -w net.ipv4.ip_nonlocal_bind=1
Once the option is set, pound can bind to the appropriate HA-ips even if they
aren't on a local interface. Which means both pound instances have an
identical configuration, and there's no need to use poundctl to manipulate it
in a failover event.
Then use pacemaker/corosync to maintain the HA-ip on the appropriate server.
If you have session affinity you'll want to use BackendCookies (or backend
session replication), or a failover event will cause session restarts.
Joe
> -----Original Message-----
> From: [email protected] [mailto:[email protected]]
> Sent: Thursday, April 19, 2012 9:14 AM
> To: [email protected]
> Subject: [Pound Mailing List] pound with heartbeat/pacemaker
>
> Hi,
>
> I´m planing to implement a ha-loadbalancer solution depending on pound
> with a heartbeat/pacemaker cluster. I think there a two possible ways
> to achive this:
>
> 1. let pound running all the time on both nodes and switch only the
> HA- IP´s between the two member.
> 2.let pacemaker start and stop the pound-process and of course let
> pacemaker handle the HA-IP´s.
>
> Question about part 1: Can pound handle added IP´s after the process
> was started ?
> Question about part 2: Is there an agent available or is there any
> other solution to start the process via pacemaker ?
>
> Any suggestions or further information is welcome
>
> Kind regards
>
> fatcharly
>
> --
> Empfehlen Sie GMX DSL Ihren Freunden und Bekannten und wir belohnen
> Sie mit bis zu 50,- Euro! https://freundschaftswerbung.gmx.de
>
> --
> To unsubscribe send an email with subject unsubscribe to
> [email protected].
> Please contact [email protected] for questions.
N r zǧu ޙ +a y n ˛ m h u l !>W ( ֜ ,z + + 笶*'
