How to Setup GFS2 or GFS in Linux Centos

It has been a nightmare for me setting up GFS2 with my 3 shared hosting servers and 1 SAN Storage. I have been reading all over the internet and the solutions to this is either outdated or contains bug that cannot make my SAN storage SAN to work. Finally, i managed to setup my GFS2 on my Dell MD3200i with 10TB of disk space.

GFS2/GFS Test Environment

Here is the test environment equipment that i utilized for this setup.

  1. 3 Centos Web Server
  2. 1 MD3200i Dell SAN Storage
  3. 1 Switch to connect all these equipment together

Assumption

I will assume you would have setup all your 3 Centos servers to communicate with your SAN ISCSI storage. This means that all your 3 Centos servers will be able to view your newly created LUN using iscsiadmn. And you have switch off your iptabls and selinux. If your iscsi storage hasn't configure, you can do so at cyberciti.

Setup GFS2/GFS packages

On all of your 3 Centos servers, you must install the following packages:

  1. cman
  2. gfs-utils
  3. kmod-dlm
  4. modcluster
  5. ricci
  6. luci
  7. cluster-snmp
  8. iscsi-initiator-utils
  9. openais
  10. oddjobs
  11. rgmanager

Or you can simple type the following yum on all 3 Centos machine

yum install -y cman gfs-utils kmod-gfs kmod-dlm modcluster ricci luci cluster-snmp iscsi-initiator-utils openais oddjob rgmanager

Or even simplier, you can just add the cluster group via the following line

yum groupinstall -y Clustering
yum groupinstall -y "Storage Cluster"

Oh, remember to update your Centos before proceeding to do all of the above.

yum -y check-update
yum -y update

After you have done all of the above, you should have all the packages available to setup GFS2/GFS on all your 3 Centos machine.

Configuring GFS2/GFS Cluster on Centos

Once you have your required centos packages installed, you would need to setup your Centos machine. Firstly, you would need to setup all your hosts file with all 3 servers machine name. Hence, i appended all my 3 servers machine name across and in each machine i would have the following additional line in my /etc/hosts file.

111.111.111.1 gfs1.hungred.com
111.111.111.2 gfs2.hungred.com
111.111.111.3 gfs3.hungred.com

where *.hungred.com is each machine name and the ip beside it are the machine ip addresses which allows each of them to communicate with each other by using the ip stated there.

Next, we will need to setup the cluster configuration of the server. On each machine, you will need to execute the following instruction to create a proper cluster configuration on each Centos machine.

ccs_tool create HungredCluster
ccs_tool addfence -C node1_ipmi fence_ipmilan ipaddr=111.111.111.1 login=root passwd=machine_1_password
ccs_tool addfence -C node2_ipmi fence_ipmilan ipaddr=111.111.111.2 login=root passwd=machine_2_password
ccs_tool addfence -C node3_ipmi fence_ipmilan ipaddr=111.111.111.3 login=root passwd=machine_3_password

ccs_tool addnode -C gfs1.hungred.com -n 1 -v 1 -f node1_ipmi
ccs_tool addnode -C gfs2.hungred.com -n 2 -v 1 -f node2_ipmi
ccs_tool addnode -C gfs3.hungred.com -n 3 -v 1 -f node3_ipmi

Next, you will need to start cman.

service cman start
service rgmanager start

cman should starts without any error. If you have any error while starting cman, your GFS2/GFS will not work. If everything works fine, you should see the following when you type the command as shown below,

[root@localhost ]# cman_tool nodes
10.0.0.1
Node  Sts   Inc   Joined               Name
   1   M     16   2011-1-06 02:30:27  gfs1.hungred.com
   2   M     20   2011-1-06 02:30:02  gfs2.hungred.com
   3   M     24   2011-1-06 02:36:01  gfs3.hungred.com

If the above shows, this means that you have properly setup your GFS2 cluster. Next we will need to setup GFS2!

Setting up GFS2/GFS on Centos

You will need to start the following services.

  • service gfs start
  • service gfs2 start

Once, this two has been started. All you need to do is to partition your SAN storage LUN. If you want to use GFS2, partition it with gfs2

/sbin/mkfs.gfs2 -j 10 -p lock_dlm -t HungredCluster:GFS /dev/sdb

Likewise, if you like to use gfs, just change it to gfs instead of gfs2

/sbin/mkfs.gfs -j 10 -p lock_dlm -t HungredCluster:GFS /dev/sdb

A little explanation here. HungredCluster is the one we created while we were setup out GFS2 Cluster. /dev/sdb is the SAN storage lun space which was discovered using iscsiadm. -j 10 is the number of journals. each machine within the cluster will require 1 cluster. Therefore, it is good to determine the number of machine you will place into this cluster. -p lock_dlm is the lock type we will be using. There are other 2 more types beside lock_dlm which you can search online.

P.S: All of the servers that will belong to the GFS cluster will need to be located in the same VLAN. Contact support if you need assistance regarding this.
If you are only configuring two servers in the cluster, you will need to manually edit the file /etc/cluster/cluster.conf file on each server. After the tag, add the following text:

If you do not make this change, the servers will not be able to establish a quorum and will refuse to cluster by design.

Setup GFS2/GFS run on startup

Key the following to ensure that GFS2/GFS starts everytime the system reboot.

chkconfig gfs on
chkconfig gfs2 on
chkconfig clvmd on //if you are using lvm
chkconfig cman on
chkconfig iscsi on
chkconfig acpid off
chkconfig rgmanager on
echo "/dev/sdb /home gfs2 defaults,noatime,nodiratime 0 0" >>/etc/fstab
mount /dev/sdb

Once this is done, your GFS2/GFS will have mount on your system to /home. You can check whether it works using the following command.

[root@localhost ~]# df -h

You should now be able to create files on one of the nodes in the cluster, and have the files appear right away on all the other nodes in the cluster.

Optimize clvmd

We can try to optimize lvmd to control the type of locking lvmd is using.

vi /etc/clvmd/clvmd.conf
find the below variables and change it to the variable as shown below
locking_type = 3
fallback_to_local_locking = 0
service clvmd restart

credit goes to http://pbraun.nethence.com/doc/filesystems/gfs2.html

Optimize GFS2/GFS

There are a few ways to optimize your gfs file system. Here are some of them.
Set your plock rate to unlimited and ownership to 1 in /etc/cluster/cluster.conf


Set noatime and nodiratime in your fstab.

echo "/dev/sdb /home gfs2 defaults,noatime,nodiratime 0 0" >>/etc/fstab

lastly, we can tune gfs directy by decreasing how often GFS2 demotes its locks via this method.

echo "
gfs2_tool settune /GFS glock_purge 50
gfs2_tool settune /GFS scand_secs 5
gfs2_tool settune /GFS demote_secs 20
gfs2_tool settune /GFS quota_account 0
gfs2_tool settune /GFS statfs_fast 1
gfs2_tool settune /GFS statfs_slots 128
" >> /etc/rc.local

credit goes to linuxdynasty.

iptables and gfs2/gfs port

If you wish to have iptables remain active, you will need to open up the following ports.

-A INPUT -i 10.10.10.200 -m state --state NEW -p udp -s 10.10.10.0/24 -d 10.10.10.0/24 --dport 5404, 5405 -j ACCEPT
-A INPUT -i 10.10.10.200 -m state --state NEW -m multiport -p tcp -s 10.10.10.0/24 -d 10.10.10.0/24 --dports 8084 -j ACCEPT
-A INPUT -i 10.10.10.200 -m state --state NEW -m multiport -p tcp -s 10.10.10.0/24 -d 10.10.10.0/24 --dports 11111 -j ACCEPT
-A INPUT -i 10.10.10.200 -m state --state NEW -m multiport -p tcp -s 10.10.10.0/24 -d 10.10.10.0/24 --dports 14567 -j ACCEPT
-A INPUT -i 10.10.10.200 -m state --state NEW -m multiport -p tcp -s 10.10.10.0/24 -d 10.10.10.0/24 --dports 16851 -j ACCEPT
-A INPUT -i 10.10.10.200 -m state --state NEW -m multiport -p tcp -s 10.10.10.0/24 -d 10.10.10.0/24 --dports 21064 -j ACCEPT
-A INPUT -i 10.10.10.200 -m state --state NEW -m multiport -p tcp -s 10.10.10.0/24 -d 10.10.10.0/24 --dports 41966,41967,41968,41969 -j ACCEPT
-A INPUT -i 10.10.10.200 -m state --state NEW -m multiport -p tcp -s 10.10.10.0/24 -d 10.10.10.0/24 --dports 50006,50008,50009 -j ACCEPT
-A INPUT -i 10.10.10.200 -m state --state NEW -m multiport -p udp -s 10.10.10.0/24 -d 10.10.10.0/24 --dports 50007 -j ACCEPT

Once these ports are open on your iptables, your cman should be able to restart properly without getting start either on fencing or cman starting point. Good Luck!

Troubleshooting

You might face some problem setting up GFS2 or GFS. Here are some of them which might be of some help

CMAN fencing failed

You get something like the following when you start your cman

Starting cluster:
Loading modules... done
Mounting configfs... done
Starting ccsd... done
Starting cman... done
Starting daemons... done
Starting fencing... failed

One of the possibility that is causing this is that your gfs2 has already been mounted to a drive. Hence, fencing failed. Try to unmount it and start it again.

mount.gfs2 error

if you are getting the following error

mount.gfs2: can't connect to gfs_controld: Connection refused

you need to try to start the cman service

Reference