Other Double-HAST

Hi,

since HAST goes master-slave on block level I am wondering, whether it is possible to get a 3-way-setup like that:

Code:
   zvol1 <-- hast-3way -->  zvol2
     |                         |
da1--+       da1 <--- hast-da1 +---> da1
     |                         |
da2--+       da2 <--- hast-da2 +---> da2
node1        node2                   node3

Disks da1, da2 in nodes 2 and 3 are merged with HAST to a zvol2.
On node 1 create simply zvol1 using da1, da2 and put on top of it HAST and merge zvol1 and zvol2 with HAST to a 3-way-2-disk-setup.

If this works, sync between zvol1 and zvol2 might be slower than of disks of hast-da1 and hast-da2. But hey, it is 3-way then ?
 
I think this is technically possible. The startup scripts could get hairy. Only the HAST primary node will gain access to /dev/hast/hast-resource-name. Meaning, the primary HAST node will, through some zvol hops, end up with two secondaries. I.E. the first secondary can't mirror to the second secondary. Cool idea to throw some Raspberry Pis/VMs at and see what happens/what kind of overhead this entails. Report back, please!
 
I think this is technically possible. The startup scripts could get hairy. Only the HAST primary node will gain access to /dev/hast/hast-resource-name. Meaning, the primary HAST node will, through some zvol hops, end up with two secondaries. I.E. the first secondary can't mirror to the second secondary. Cool idea to throw some Raspberry Pis/VMs at and see what happens/what kind of overhead this entails. Report back, please!
actually node 1 is master of hast-3way, node 2 is master for hast-da{0,1}, but slave for zvol2, node 3 is double-slave OR node 3 is master for hast-da{0,1}, but slave for zvol2, node 2 is double-slave.
nevermind of any failover script ?

I will try it by the end of the week is spare time left over ...
 
Back
Top