Dear all ,
I am facing the problem of read only filesystem with Heart beat configuration.
Here are my setup details :--
=========================
I have installed the 2 physical server RHEL 5.2 64 bit with Xen virtualization
and IBM BS 8300 SAN. The XEN kernel verison is 2.6.18-92.el5xen. The File
system is EXT3.
I am using this below setup for a heavy mail server where the POP3 and IMAP
traffic is very high ( 1.2 lack emails per day) , I am using cyrus and postfix
and ldap
Each physical server conatin XEN vitual OS. Outof this one is master and other
slave for me.
The physical server1 IP :-- 192.168.1.1
The physical server2 IP :-- 192.168.1.2
The Xen VM machine IP under physical server1 :-- 192.168.1.10 ( Master )
The Xen VM machine IP under physical server2 :-- 192.168.1.20 ( Salve )
The HA folting IP between 2 VM is 192.168.1.30
Both the VM and physical machine communicate to each other via switch and not
by cross cable.
Inside the both Xen VM I am using the 4 SAN partition ( for email boxs), which
is accessable and can be mount from 192.168.1.10 & 192.168.1.20.
I am managing the mounting and unmounting of the SAN partition and stoping and
starting the services vai script , which is mention in /etc/ha.d/haresources
My problem :--
============
1) When 192.168.1.10 is master with flowting IP 192.168.1.30 , at that time
every thing work proper.
But some time due to n/w problem the flowting IP 192.168.1.30 get switch to
slave server i.e 192.168.1.20. At that time while mouting the SAN partiton ,
the san parititon goes in readonly mode , and for correction this require to do
fsck -y <device deriver>
i.e failover causing the filesystem to go in read only mode.
I observer that this read only problem not come everytime , but it happen it
messup evrything.
2) When both the server get lost there n/w connection , both assume that there
respective slave/master is down and they both server act as master-master
server and also mount the SAN partition on each VM server.
So ,
1) How to avoid the redaonly file system problem ?
2) How to avoiod master-master problem , in case of n/w failuer.
3) What kindof precuosion I should take while doing mounting and umounting SAN
prartition. ?
I tried Ipfail and ping in ha.cf but no luck.
I am using Heartbeat version 2.1
heartbeat-2.1.4-9.el5.x86_64.rpm
heartbeat-pils-2.1.4-9.el5.x86_64.rpm
heartbeat-stonith-2.1.4-9.el5.x86_64.rpm
Here is my ha.cf (same on both active passive server ):--
===========================================
cat /etc/ha.d/ha.cf
debugfile /var/halogs/ha-debug
logfile /var/halogs/ha-log
logfacility local0
keepalive 2
deadtime 15
warntime 10
initdead 30
udpport 694
bcast eth0
auto_failback off
node activeimap1
node passiveimap1
#ping 192.168.2.8
#respawn hacluster /usr/lib/heartbeat/ipfail
debug 0
Here is my haresource (same on both active passive server ) :--
================================================
cat /etc/ha.d/haresources
activeimap1 IPaddr::192.168.1.30 ms4-services
Here is my ms4service (same on both active passive server ) :--
==================================================
cat /etc/ha.d/resource.d/ms4-services
#!/bin/bash
set -x
mylist="syslog
crond
ldap
postfix
saslauthd
cyrus-imapd
httpd
"
stop() {
for i in $mylist; do
/etc/init.d/$i stop
/etc/init.d/$i stop
/etc/init.d/$i stop
done
/sbin/ifdown eth0:2
/sbin/ifdown eth0:3
/sbin/ifdown eth0:4
/sbin/ifdown eth0:5
if
cd /tmp/
/bin/mount | grep -e " on /usr/local " >/dev/null
then
# Kill all processes open on filesystem
/sbin/fuser -muk /usr/local
/sbin/fuser -muk /imap
/sbin/fuser -muk /imap1
/sbin/fuser -muk /imap2
sleep 3
/bin/umount /dev/xvdb
sleep 3
/bin/umount /dev/xvdd
sleep 3
/bin/umount /dev/xvdc
sleep 3
/bin/umount /dev/xvdf
fi
}
start() {
sleep 150
/bin/mount /dev/xvdf /usr/local
sleep 3
mount /dev/xvdc /imap1
sleep 3
mount /dev/xvdd /imap
sleep 3
mount /dev/xvdb /imap2
sleep 3
/sbin/ifup eth0:2
/sbin/ifup eth0:3
/sbin/ifup eth0:4
/sbin/ifup eth0:5
for i in $mylist; do
/etc/init.d/$i start
done
}
case $1 in
start)
start
;;
stop)
stop
;;
*)
echo "You passed wrong option {start|stop}"
exit 1
;;
esac
Please guide
Thanks and Regards
Jayesh Shinde
_______________________________________________
Linux-HA mailing list
[email protected]
http://lists.linux-ha.org/mailman/listinfo/linux-ha
See also: http://linux-ha.org/ReportingProblems
