[Pacemaker] Assessment of the performance of a Pacemaker/Corosync active/active cluster

Andrew Beekhof andrew at beekhof.net
Tue Apr 20 07:32:50 UTC 2010


On Mon, Apr 19, 2010 at 1:40 PM,  <fpetretti at secure-edge.com> wrote:
>> On Fri, Apr 16, 2010 at 3:05 PM, Francesco Petretti
>> <fpetretti at secure-edge.com> wrote:
>>> We have a couple of VMWare virtual machines, both running the same
>>> Operating
>>> System (RHEL 5.4);
>>> these guest systems run on different virtualization hosts.
>>>
>>> Both nodes run an Apache web server, serving the pages: "I am node 1"
>>> and "I
>>> am node 2".
>>>
>>> We have noticed that if a client accesses the floating IP, we get a
>>> transactions per second (TPS) number which is approximately 20% of the
>>> number that we get contacting real IPs.
>>>
>>> The configuration of these two nodes was done as outlined in:
>>> http://www.clusterlabs.org/doc/Cluster_from_Scratch.pdf
>>
>> Which step are you up to?
>> Has the IP been cloned?
>>
> Well I think I have cloned it:

This:

> clone WebIP ClusterIP \
>        meta globally-unique="true" clone-max="2" clone-node-max="1"

means you have :-)

Basically the slowdown is coming from iptables:

		$IPTABLES -I INPUT -d $BASEIP -i $NIC -j CLUSTERIP \
				--new \
				--clustermac $IF_MAC \
				--total-nodes $IP_INC_GLOBAL \
				--local-node $IP_INC_NO \
				--hashmode $IP_CIP_HASH

Maybe there are some changes that could be made here to speed things
up (I'm not really an iptables expert).
Its really just a poorman's load balancer, if the performance isn't
enough you might have to look at a hardware based solution that sits
in front of the cluster.

>
> [root at pes-app-sviluppo-05 ~]# crm
> crm(live)# configure show
> INFO: building help index
> node pes-app-sviluppo-04
> node pes-app-sviluppo-05
> primitive ClusterIP ocf:heartbeat:IPaddr2 \
>        params ip="192.168.41.238" cidr_netmask="32"
> clusterip_hash="sourceip-sourceport" \
>        op monitor interval="30s"
> primitive WebSite ocf:heartbeat:apache \
>        params configfile="/etc/httpd/conf/httpd.conf" \
>        op monitor interval="1min"
> clone WebIP ClusterIP \
>        meta globally-unique="true" clone-max="2" clone-node-max="1"
> clone WebSiteClone WebSite
> colocation website-with-ip inf: WebSiteClone WebIP
> order apache-after-ip inf: WebIP WebSiteClone
> property $id="cib-bootstrap-options" \
>        dc-version="1.0.8-9881a7350d6182bae9e8e557cf20a3cc5dac3ee7" \
>        cluster-infrastructure="openais" \
>        expected-quorum-votes="2" \
>        stonith-enabled="false" \
>        no-quorum-policy="ignore"
> crm(live)#
>
>
> ============
> Last updated: Mon Apr 19 13:33:46 2010
> Stack: openais
> Current DC: pes-app-sviluppo-04 - partition with quorum
> Version: 1.0.8-9881a7350d6182bae9e8e557cf20a3cc5dac3ee7
> 2 Nodes configured, 2 expected votes
> 2 Resources configured.
> ============
>
> Online: [ pes-app-sviluppo-04 pes-app-sviluppo-05 ]
>
>  Clone Set: WebIP (unique)
>     ClusterIP:0        (ocf::heartbeat:IPaddr2):       Started pes-app-sviluppo-04
>     ClusterIP:1        (ocf::heartbeat:IPaddr2):       Started pes-app-sviluppo-05
>  Clone Set: WebSiteClone
>     Started: [ pes-app-sviluppo-04 pes-app-sviluppo-05 ]
>
>
>
>>> Which parameters should we modify to improve the performance? (if the
>>> performance _can _be improved...).
>>> Can performance be improved ?
>>>
>>> --
>>> /******
>>> FP a.k.a. bip
>>> *****/
>>>
>>>
>>> _______________________________________________
>>> Pacemaker mailing list: Pacemaker at oss.clusterlabs.org
>>> http://oss.clusterlabs.org/mailman/listinfo/pacemaker
>>>
>>> Project Home: http://www.clusterlabs.org
>>> Getting started: http://www.clusterlabs.org/doc/Cluster_from_Scratch.pdf
>>>
>>
>
>
>




More information about the Pacemaker mailing list