I built a test cluster with SLES11SP4 (on Power). Resource group is simple: LVM mirror that resides on several LUNs from two different storages and filesystem (xfs) on top of it. Stonith resources are also configured. clvmd and dlm are up and running. No problem in configuration so far.
What is strange: if I decide to test resource transfer and reboot node (current resource owner) using shutdown -r now or stop it using halt (which is pretty much the same in Linux), I end up in node hang. In the same time resource group is moved out to another node without any issues and again, no problem here.
After a little investigation I stumbled upon hanging "vgchange -an" process that was executed from /etc/init.d/boot.d/K05boot.lvm.
So the questions are:
Is it wrong or is it intended behaviour? I understand that in my case volume group may be opened on one node only and I second it, but infinite loop in vgchange doesn't seem right even if clvmd is not available by the moment. Am I doing something wrong?