XEN migration issue using GFS2
Hi all,
I have two nodes cluster running on RHEL52 which are up to date.
I have shared disk device connected to nodes and his size is 15G.
I created GFS2 file system using DLM_lock and two journals.
Now I can mount this device using my cluster services and use this disk for sharing of Virtual Machine image file (size almost 15G).
The issue is:
- The device is mounted on both nodes.
- I can start/stop VM guest on one node properly and repeate this anytime
- Issue is when I stop VM on one node and try to start it on other one. The starting process (creation of VM) is freeze and I cannot perform any I/O operation on mointpoint (like list of folder etc.) I cannot perform mountpoint on this node because device is busy
- It seems like there is some lock hanged up on the first node and the second one is not able to write to the image file.
- The situation is same in case that I am do migration using rgmanager cluster service.
This issue occurs when I have device mounted on both nodes during all time. When I start/stop the VM guest on the first node and then I perform umount and mount again. The second node is able to start VM guest.
Is there any way how to determine which lock is active or hang on the first node? Is this a bug of GFS2 or is there any workaround on this.
Could someone help me on this. Is there some workaround?
Thank you veru much.
regards,
Stefan
|