fence_scsi howto

[csb1582]

Hi, cool project, thanks!

I have developed a FlexVolume plugin that manages clustered shared disks. The plugin fences the SCSI disks with persistent reservations via sg_persist while the volume is in use. When volume dismounted the SCSI lock is released. Seems to work well, except in case of node failure, when lock is never released (because proper teardown didn't happen.)

Do you know if it's possible to use this project to STONITH a failed node, and then remove the SCSI reservation on shared disk from another node?

[kvaps]

Hey, sorry for long answer, try to create new pod template with script for find and removing SCSI reservation. And specify it in fencing/after-hook annotation for your fencing pod template.