Openfiler DRBD + HA

The scenario I will try to illustrate, is 2 Openfiler systems in alinux-ha environment. For simplicity we shall refer to these systems asB1 and B2. B2 is configured to work as a hot standby, constantlychecking if B1 is up. If anything happens to B1, B2 comes up, takingover the services of B1.
From the outside world, only one system is visible at any point intime. This is done via a virtual ip address that is always assigned toone server. If anything happens to B1, and B2 decides to take over, itsends an ARP broadcast to the switch saying, hey, B1’s mac just changes,I am B1. More about this can be read at http://www.linux-ha.org
These are the steps I have outlined in my notes to get things going(these are bare instructions, please read the fine manuals):
1.) Prepare your hardware. My systems have 3 Nics. 1 10/100 for theWAN, 1 Gigabit card for DRBD sync, 1 10/100 for running daily backups ofthe system files. DRBD is what will replicate your volumes. I used acrossover cable between the two cards, but you can always use a switch.
2.) Install Openfiler on both systems, do not do any LVM setups inDiskdruid, just leave the free space as is, you will come back to itlater manually.

2a.) Setup ip networks, naturally separate on each NIC.

2b.) Most of you will have your WAN turned off, but if for any reasonyou need your WAN NICs on, make sure your iptables rules are set.

Once all of your system is setup, and you can communicate on allinterfaces, you can go on to setting up DRBD.
3.) DRBD is setup from one conf file, /etc/drbd.conf. Here is anexample:##############resource r0 { protocol C; incon-degr-cmd “echo ‘!DRBD! pri on incon-degr’ | wall ; sleep60 ; halt -f”; startup { wfc-timeout 10; degr-wfc-timeout 120; } disk { on-io-error detach; } net { timeout 60; connect-int 10; ping-int 10; #max-buffers 2048; max-epoch-size 2048; } syncer { rate 100M; group 1; #al-extents 257; } on B1 { device /dev/drbd0; disk /dev/hda4; address 192.168.30.30:7788; meta-disk internal; } on B2 { device /dev/drbd0; disk /dev/hda4; address 192.168.30.31:7788; meta-disk internal; }}#########################
All of these config options are described in detail in the originalinstall file. Please also pay attention to the version of DRBD you haveinstalled (rpm -qa | grep drbd). Configs have changed since 0.6.
Once you have been successfull at configuring both sides, start drbdvia /etc/rc.d/init.d/drbd.
You can watch the status of your sync via “cat /proc/drbd”. That willtell you what is the status of DRBD. Most likely you will be inSecondary/Secondary mode, which means that neither one of B1 or B2 isable to access the share. Remember, only one side can be primary at atime.
Drbdadm is the tool to control the state of your DRBD. By issuing thefollowing command :
“drbdadm primary r0”, you will bring B1 or B2 to primary mode. Oncethe state is considered primary, you can start accessing it from thatserver.
You should practice setting your share to primary and secondary stateon both sides. Issue the following commands:
B1: drbdadm secondary r0
B2: drbdadm primary r0
B2: cat /proc/drbd
B2: drbdadm secondary r0
B2: cat /proc/drbd
B1: cat /proc/drbd
B1: drbdadm primary r0
B1 cat /proc/drbd
If you have been successful, you should see the state changes from oneside to another.
Now, lets go on to creating LVM volumes and groups.
4.) Bring both B1 and B2 to state Sec/Sec (DRBD)
LVM HOWTO: http://www.tldp.org/HOWTO/LVM-HOWTO/index.html
4a.
Bring B1 into primary
4b.
pvcreate /dev/drbd0

If you say /dev/hda* you are shooting yourself in the foot.
4c.
vgcreate openfiler /dev/drbd0

activate via vgchange -a y openfiler
Now, you should be able to visit http://blabla:446 login, click volumesand vuala. You should see your volume. Do not create anything at thistime.
4d.
Put B1 in secondary.

deactivate lvm, vgchange -a n openfiler
4e.
Put B2 in primary.
4f.
pvcreate /dev/drbd0
4g.
vgcreate openfiler /dev/drbd0

activate lvm, vgchange -a y openfiler

You should be able to see your volume on B2 now (via the webinterface).

Now here is what I put into my .bashrc file on both systems:alias prim=”drbdadm primary r0 && vgchange -a y openfiler”alias sec=”vgchange -a n openfiler && drbdadm secondary r0″alias disp=”cat /proc/drbd”
This way you can just type “disp” to see the status, and the logicbehind failing over is this:When you execute “sec”, your volume group deactivates and drbd goessecondary after that to ensure no writing. On the other side, when youissue “prim”, the DRBD network drive comes up to primary, then thevolume group gets activated.Now, when you create your volume in Openfiler, you will see that you canno longer “go secondary”, it will say something like : vgchange — can’t deactivate volume group “openfiler” with 1 openlogical volumeI have the question into LVM right now, I will update you guysfurther. Also need to figure our how to setup heartbeat, but I hope the abovecleared up some gray areas some of you could have had.I am terrible at writing manuals, so if you want help, come to the IRCchan on FreeNode, #openfiler, I should be able to help!

Notes

Leave a Reply

Your email address will not be published. Required fields are marked *