Running SuSE Linux Cluster on VMware
Posted by jlchannel - on September 12th, 2008 in Operating Systems, Tips | 15 Comments »
What? Running SuSE Linux Cluster on VMware ESX? Basically if you have VMware HA and DRS enable, your virtual machine is running on cluster mode. Then, why running Linux cluster on VMware? Who care? It’s just for fun and for personal testing purpose only.
1st, create two SuSE Linux(I’m running SLES 10 SP2) on VMware running on same network said cluster1(192.168.1.1) and cluster2(192.168.1.2).
2nd, add new hard disk on VMware infrastructure client. Please remember to choose ‘Use an existing virtual disk‘.
Click here for bigger image.
3rd, install and configure Linux cluster heartbeat.
Make sure you have heartbeat RPM install
cluster1:~ # rpm -qa|grep heartbeat
sles-heartbeat_en-10.1-0.20
heartbeat-pils-2.1.3-0.9
heartbeat-stonith-2.1.3-0.9
yast2-heartbeat-2.13.13-0.3
heartbeat-2.1.3-0.9
cluster2:~ # rpm -qa|grep heartbeat
sles-heartbeat_en-10.1-0.20
heartbeat-pils-2.1.3-0.9
heartbeat-stonith-2.1.3-0.9
yast2-heartbeat-2.13.13-0.3
heartbeat-2.1.3-0.9
Install Linux cluster heartbeat RPM package if you haven’t install
cluster1:~ # rpm -ivh heartbeat-xxxxxx
cluster1:~ # rpm -ivh heartbeat-stonith-xxxxxx
and etc
Edit Authentication file (same for cluster1 and cluster2)
cluster1:~ #vi /etc/ha.d/authkeys
auth 2
2 sha1 MVM_CLUS2!
Edit HA configuration file
cluster1
cluster1:~ #vi /etc/ha.d/ha.cf
logfile /var/log/ha-log
logfacility local0
node cluster1
node cluster2
keepalive 1
warntime 30
deadtime 60
initdead 120
udpport 694
ucast eth0 192.168.1.1
ucast eth0 192.168.1.2
auto_failback off
ping_group group1 192.168.1.254
respawn hacluster /usr/lib/heartbeat/ipfail
watchdog /dev/watchdog
cluster2
cluster2:~ #vi /etc/ha.d/ha.cf
logfile /var/log/ha-log
logfacility local0
node cluster2
node cluster1
keepalive 1
warntime 30
deadtime 60
initdead 120
udpport 694
ucast eth0 192.168.1.2
ucast eth0 192.168.1.1
auto_failback off
ping_group group1 192.168.1.254
respawn hacluster /usr/lib/heartbeat/ipfail
watchdog /dev/watchdog
Edit HA Resource file (same for cluster1 and cluster2)
cluster1:~ #vi /etc/ha.d/haresources
cluster1 192.168.1.100 \
Filesystem::/dev/sdb1::/db::ext3
Make sure you can see hard disk partition on both servers
cluster1:~ # cat /proc/partitions
8 17 20964793 sdb1
cluster2:~ # cat /proc/partitions
8 17 20964793 sdb1
Configure and format hard disk partition
cluster1:~ # fdisk /dev/sdb1
press n add a new partition
press p primary partition
press 1 partition number
First cylinder, press ENTER
Last cylinder or +size or +sizeM or +sizeK, press ENTER
press p print the partition table
press w write table to disk and exit
cluster1:~ # mkfs.ext3 /dev/sdb1
Create new folder on both servers
cluster1:~ # mkdir /db
cluster2:~ # mkdir /db
Startup heartbeat services
cluster1:~ # chkconfig heartbeat on
cluster1:~ # /etc/init.d/heartbeat start
cluster2:~ # chkconfig heartbeat on
cluster2:~ # /etc/init.d/heartbeat start
You should be able to see /db mounted and IP address 192.168.1.100 is up on cluster1(active). If you turn OFF cluster1 server and you should see cluster2 running on ACTIVE mode.
Congratulations! You’re done! Enjoy!
15 Responses
Why don’t you come and post this over on VIOPS in our Availability zone?
Thanks for your suggestion Steve.
in my case , the customer Not willing to purchase the SAN or iSCSI , and the only way I can deploy the vm with HA is by using linux-ha , but I found out if I use broadcast , sometime the virtual ip is not pingable , if I move to unicast , it solve the issue , I also notice I need to use sha1 , if I using crc , there will be some issues .thanks !
I will suggest you try some open source for shared storage example like open filer, software ISCSI and NFS
Probably you may check this Setup a Linux Highly Availability NFS servers
http://www.planetmy.com/blog/setup-a-linux-highly-availability-nfs-servers/
recently I put heartbeat on vmware ESX 3.5 U2 standard , totally 15 VM on each ESX ,
I have some weird issue , the heartbeat seem restart itself occasionally and cause cluster partition happen , I am not sure if heartbeat really work in vmware
do you meant the hearbeat on VM guests clustering or ESX host clustering?
sorry for confuse , I mean running linux heartbeat on VM guests ( SLES9 SP3 ) the heartbeat version is 1.2.3 , from SLES9 repository , but I upgrade to 2.1.3 ( from source ) , still encounter very inconsistent on the Linux guests .
root cause found , the hardware I use have share interrupts on raid controller and NIC , cause the bottleneck on the network . I check the /proc/vmware/interrupts how that they shared same IRQ .
Hi All, we have the forum created and welcome to register as a member and post more question in our forum.
thanks
if anybody can help with four node architecture using heartbeat problem is how to stablize ip t second node if first goes down as floating between third and then settlin on second causing data loss any script or Ha.cf parameter can solve this
Thanks in advance
Will the same configuration work for CentOS Ver.5.5 distribution.
Maddy, I’m not sure. Please let me know the result if you have time to test it.
Ye sure… I will going to test it on CentOS ver. 5.5 and definately I will post the result…
CentOS 5 uses luci/ricci out of the box. Not very well though. CentOS 6 made a lot of improvements on cluster. It is harder to set up with quorum disk but worth it if you take time to get timings right. Otherwise, SLES (and I think OpenSUSE) works well with pacemaker/openais. Much easier to manage. I did a write-up of outline of steps for SLES, I can do one on CentOS 6 soon too if anyone is interested.
http://geekswing.com/geek/building-a-two-node-sles11-sp2-linux-cluster-on-vmware/