Download as txt, pdf, or txt
Download as txt, pdf, or txt
You are on page 1of 1

##Intro

PCS (Pacemaker Cluster Services) provides high availability services for


CentOS/RHEL 6+.

##Pre-Reqs
/etc/hosts should contain aliases to the PROD IPs with a -cl suffix (to match
Matt's auto iptables script) e.g.

10.11.12.13 node1-prod node1-cl


10.11.12.14 node2-prod node2-cl
Following ports should be opened between nodes

2224/tcp pcsd (ruby)


5405/udp corosync

##Install PCS Cluster


# Both MCs as root, install and start PCS, configure hacluster user
yum install pcs corosync pacemaker
service pcsd start && chkconfig pcsd on
passwd hacluster # hacluster, on a prod system this should be something secure!

# MC1 only as root, auth cluster nodes and start cluster


pcs cluster auth node1-cl node2-cl # enter hacluster user/pass from above passwd
command
pcs cluster setup --name sfwmc node1-cl node2-cl
pcs cluster status
pcs cluster start --all
pcs cluster status
pcs status

# Both MCs as root, configure cluster


corosync-cfgtool -s
pcs status corosync
journalctl | grep error
crm_verify -L -V
pcs property set stonith-enabled=false
crm_verify -L
pcs resource defaults resource-stickiness=100
pcs cluster enable --all
crm_verify -L -V

##Gotchas
pcsd.service start operation timed out. Terminating.

netstat -anp |grep ruby # showed a number of connections to 8.8.8.8:53 (DNS)


Turns out /etc/resolv.conf had unnecessary entries

cat /dev/null >/etc/resolv.conf

You might also like