Comments on: DRBD Primary/Primary using GFS

By: Shaker

Shaker — Wed, 07 Oct 2009 06:29:11 +0000

Great, you are welcome ,,
No, i am a sysadmin not network guy

By: ilyas

ilyas — Tue, 06 Oct 2009 23:19:52 +0000

ok, i become clear.. thanks for the suggestions.. Are u a network administrator..? Have a nice job Shaik!! and share many nice documents more..
thanks for all!

By: Shaker

Shaker — Tue, 06 Oct 2009 07:24:35 +0000

Cong. on building your Primary/Secondary arch.

>> “but, i want all loads from the clients automatically balanced to both nodes and when one of nodes fails, the active node automatically becomes a primary one..”
what are you asking for is what this document explains, there is no Secondary at all in this setup, two nodes are Primary, but i think you have to consider many problems may appear from this setup, specially when one of the nodes became down, please read more about it.

a Clustered setup means that you need a file system that support clustering, and this why GFS is needed.

I suggest you to stay with Primary/Secondary unless you really needs Primary/Primary setup.

By: ilyas

ilyas — Tue, 06 Oct 2009 04:07:57 +0000

i’ve succeed make it in primary/secondary mode. but, i want all loads from the clients automatically balanced to both nodes and when one of nodes fails, the active node automatically becomes a primary one..

now i’m building a clustered server for my web and database server. it needs a load balancer to handle the loads from clients and synchronizing data on both nodes. i use heartbeat and drbd.. am i need a GFS service to solve my problems..?
thanks for the solutions..!

By: Shaker

Shaker — Sun, 04 Oct 2009 13:43:03 +0000

Hello ilyas,
what you are confused of? i think if you don’t really need them load balanced (your application can run in one server) then don’t make them primary and stay with Active/Passive solution.
you need a tut. about what exactly?

By: ilyas

ilyas — Fri, 02 Oct 2009 01:41:02 +0000

Hi Shaker, how about heartbeat..? i try to combine it with drbd using ip alias. but i confuse to make them primary-primay node and build a load balancer to distribute the load.. any tutorials may be..? thanks a lot..!

By: Shaker

Shaker — Thu, 30 Jul 2009 09:45:00 +0000

Adam, sorry i didn’t face this thing before, files i had in the system was between 1-10MB;

By: adam

adam — Tue, 28 Apr 2009 23:40:46 +0000

hi-
ive used this setup and doing some tests with drbd. ive noticed that when i write a large file (>500M or so) to /mnt on node1, for example, i/o to the filesystem is stalled on node2 until drbd has completed replication of the file. for instance, if i cp a file to /mnt on node1 it will write at full local I/O speed and complete the write.. however, node1′s I/o is hangng until drbd catches up with the filesystem.. is this normal? is there a way to tweak this? changing drbd protocol perhaps? this is my first setup and have only spend an hour or so testing it. curious..

By: Shaker

Shaker — Fri, 13 Feb 2009 15:29:40 +0000

it was working well for about 6 months, i remember one or two crashes happened, we recovered them just as pointed in drbd.org
for now we expanded the number of nodes, now we use 4 nodes, so drbd didn’t help us in this setup, i used GFS and GNBD from redhat to setup this, 1 GNBD server as storage, and 3 Nodes running Apache.

By: Amos Shapira

Amos Shapira — Fri, 13 Feb 2009 02:11:02 +0000

Now that you have almost a year of using this setup – how stable is it for you?

Did you have cases of one of the nodes crashing (for any reason) and recovering? What did you have to do to get it back?

Thanks.