By Jerome Blanchet on mai 27, 2009
In case of hardware Failure it is good to have a very simple way to manually restart LDoms in few seconds in one other server. The Warm and Cold Migration is a very good way for moving LDoms from a primary physical platform to a secondary physical platform. But in case of crash of primary platform the Migration could be hard to use. Also if we wish to have a simple way to easily move at any time a several LDoms to a pool of several physical servers it could be hard to maintain all vdisks configurations in all physical servers for all LDoms.
One simple solution is to create all vdisks in a share file system (NFS - PXFS) or failover file system (ZFS - metaset) and export the LDom and all associated vdisks and vswitches configurations in that File System. Then we can used this exported informations on the secondary platform to restart the LDom. Before restart the LDom on secondary platform we must check the availability of resources (CPU/MAU/Memory), services (vdiskservers/vswitches) and all backend files needed to restart the LDom. Then we can recreate all vdiskserverdevices on the right backends with right options for the specific LDom.
The fldom (Flying LDom) tool can be used to easily manage Flying Ldoms. This tools allow us to export LDom configuration in share or failover file system. Export option example:
The fldom tool allow us to detach the LDom from one server (remove it and delete all vdsdev), and attach the LDom to a new server (check needed services and resources, recreate all needed vdsdev, create and bind the LDom). Attach and detach options example:
So it is very easy to quickly restart the LDom in any server in case of primary platform failure with fldom attach option and exported informations. For example primary T5xx0 Failure scenario:
If using ZFS as failover file system for Flying LDoms without SunCluster we have to be careful of ZFS import with force option as there is no yet the support of Multi-import protection in ZFS.
The fldom tool is available here for download as simple Solaris package. Don't forget to call SunPS to successfully help you to implement this Flying LDoms solution with fldom tool. The next step will be HA-LDom to allow automatic LDom failover with SunCluster, I hope coming soon. Have fun with LDoms.