[Linux-HA] Is this the right config?
kbyrd-linuxha at memcpy.com
Mon Oct 1 08:58:22 MDT 2007
> groups can not contain clones or master/slave resources (which are
> just a special kind of clone)
> you can however clone a group
But the groups are all exact clones, they have different instance
attributes that are passed into the resource scripts, I'll explain more
>> Basically, I'm looking for how to scale this out easily without a ton of
>> XML entries for the 22 things I'm managing.
> it depends what is scaling up... the number of m/s or the number of
> items that need to run on the master
Let me back up and explain the bigger picture, that will probably help.
I have two nodes. nodeA, nodeB. Across these two, I want to run 22 "things".
A "thing" is just a name I'm using to describe a logical grouping, I want
to avoid overloading the term "group" or "resource". Let's call the
thing-1 to thing-22.
In heartbeat terms, "thing-N" is really the following in heartbeat terms:
- master/slave drbd resource, each one with a instance attribute like this:
<nvpair id="ia-drbd-N-1" name="drbd_resource" value="drbdN"/>
- Filesystem resource, with instance attributes:
<nvpair id="ia-fs-N-1" name="fstype" value="ext3"/>
<nvpair id="ia-fs-N-2" name="directory" value="/mnt/shareN"/>
<nvpair id="ia-fs-N-3" name="device" value="/dev/drbdN"/>
- A "VM" resource. I wrote the OCF resource script. It has one instance
<nvpair id="ia-vm-N-1" name="vmxfile" value="/mnt/shareN/vm.vmx"/>
In all these cases, "N" is 1 thru 22, depending on the "thing" it belongs
to. The VM resource required the Filesystem resource to be running, the
Filesystem resource requires the drbd resource to be master. Half of these
"things" will prefer to run on nodeA, half will prefer to run on nodeB,
but all 22 may run on a single node if the the other node fails.
Any help with the simplest possible cib.xml would be appreciated.
More information about the Linux-HA