This content has been marked as final. Show 7 replies
I can't see anything in the manual that states that this is not supported. However, I would expect that you would need to put both the GlobalDevicePaths and FileSystemMountPoint properties into the same resource definition so that you didn't get a conflict between the file system primary I/O path and the device group one. Could you tell me where the other source of information comes from?
Because of the thread
SC 3.3: How to reinstall second node
we opened a CASE at our supporter and the information was an additional information ( but has nothing to do with our problem in the above mentioned thread ).
Could you please advice me how the create statement should be?
clrs create -g sybase-rg -t SUNW.HAStoragePlus:9 \
-p GlobalDevicePaths=sybase \
-p FileSystemMountPoints=/xyz \
-p AffinityOn=TRUE sybase-dg
Although, having both there is probably not necessary, because moving the file system primary would bring the file system with it. I'm not totally sure that the converse is true (clearly it ought to). If you've raised a case, then the experts on this code should be able to provide a definitive answer (and solution).
I checked our create scripts and we've got extra create statements for our filesystems, like:
clrs create -t SUNW.HAStoragePlus:9 \
-g sybase-rg \
-p Resource_dependencies=sybase-dg \
-p R_description=Mount\ for\ SYBASE-Software \
-p FilesystemMountPoints=/softsybase/sybase_prod \
-p ZpoolsSearchDir= mount_sybase_prod
So I think both variants
a. put the FilesystemMountPoints and GlobalDevicePaths in one create statement
b. put them in different create statements
A short question:
About one year ago I bought a book "Oracle Solaris Cluster Essentials" from Tim Read.
Did you write this book?
OK, now I'm a little confused. You say you have a metaset and a file system which looks like a zpool. I don't think you've created the zpool on the metaset (that would not be good), so I assume you have a zfs file system and some global devices on a metaset. If so, then having two HASP resources is OK.
... and yes, I am the person who wrote the book, though as the preface points out - with a lot of input from my colleagues.
OK, my mistake. Just remove the ZpoolsSearchDir as it unnecessary anyway. Just have the filesystemmountpoint property set. That should then reference a file system that is build on a metaset which, in turn, should refer to a device group that will be migrated when the file system is switched between nodes.
I would expect that to be sufficient.