HAST recovery question

G

gnoma

Hello,

I read the entire HAST chapter in the Handbook. As far as I understand there, the configuration they gave for example does the failover switch by detecting the CARP interface state.

My question cover scenario with ZFS and iSCSI target with HAST.
  1. I am going to use ZFS for RAID management. For better performance and fault tolerance I am going to use two HAST nodes with 4 disks 1 TB each in RAID0, instead of two standalone nodes with RAIDZ. This is because HAST can give me fault tolerance for one entire storage server, RAIDZ can give me fault tolerance for only one disk in the server.
  2. I am going to create a block device using zfs tools: zfs create -b 128k -o sync=always -o compress=on -s -V 4TB zfsroot/datacore.
  3. I am going to present this block device to VMware ESXi hosts with istgt.
  4. I am willing to do a HAST failover configuration on device /dev/zvol/zfsroot/datacore and present device /dev/hast/hast-pool instead.
  5. I am going to configure CARP devices on the two2 nodes and make the iSCSI initiators look for the iSCSI target on that CARP IP.
Following the HAST chapter in the Handbook: When the master node fails, the CARP will fail over to the backup node. HAST will detect that and will also fail over. The HAST replication will be broken and here comes the important stuff - When I fix the primary node and bring it back online. The CARP will automatically switch the master interface to the primary node, the HAST will detect that and also switch the primary node too.

My question: shouldn't there be time for build/resync between the HAST nodes?

I always configure CARP advskew in case of recovery to return automatically the master interface to the primary node. But is there a ready solution that will make CARP wait until HAST resync is completed, or I have to write custom scripts for this?

Thank you.
 
You must log in or register to reply here.
Back
Top