Hi I got a san disk visible on two nodes (global or zone). On the first node, i can create a pool using "zpool create x1 sandisk". If i try to reuse this disk on the first node, i got a "vdev in use" warning. If i try to create a pool on the second node using the same disk, "zpool create x2 sandisk", it works fine, without warning, before leading to obvious problems. I am using sol10 u4 . did anyone encounter the same problem on opensolaris or s10 ? What could i be missing ? this happens whatever the NOINUSE_CHECK variable is set to. thanks a lot christophe This message posted from opensolaris.org
This is probably because ZFS is not supported as a global filesystem. If you move the zpool between cluster nodes, you''ll need to zpool export it on the first node, and zpool import it on the second node. -Tim This message posted from opensolaris.org
When moving pools, we use of course export/import or sczbt suncluster stuff. Nevertheless, we dont want to use zfs as global FS with concurrent access, just use it like svm or vxvm to declare "volumes" usable by cluster''s nodes (and used by only once at a time). so, it seems to me a bit problematic that once a LUN has been used by a storage resource declared on a node, i still can "zpool" it on the other node without warning. At least, VxVm asks for a "-f" flag. thanks for your answer, Tim. This message posted from opensolaris.org
Christophe Rolland wrote:> When moving pools, we use of course export/import or sczbt suncluster stuff. > Nevertheless, we dont want to use zfs as global FS with concurrent access, just use it like svm or vxvm to declare "volumes" usable by cluster''s nodes (and used by only once at a time). > > so, it seems to me a bit problematic that once a LUN has been used by a storage resource declared on a node, i still can "zpool" it on the other node without warning. At least, VxVm asks for a "-f" flag. >You must be running an older version of Solaris. The force (-f) flag for zpool import has been around a long time, but was not in the older Solaris 10 releases. Sorry I can''t be more specific, but I haven''t run Solaris 10 in 3 years or so. Note that it is expected that the cluster will force import, so in a sense, it doesn''t matter if the -f flag is there or not. The proper ownership and failure fencing is performed by Solaris Cluster, not ZFS. -- richard
> Note that it is expected that the cluster will force > import, so in ai was talking about creation, not import.> You must be running an older version of Solaris. Thes10u4 + sc 3.2 Anyway, bug has now been accepted. With cluster and SAN, zfs does _yet_ not behave normally :) Thanks for your answer. This message posted from opensolaris.org