Campus cluster for services inside LDOMs

Question

Hi,
Is it possible to set up campus cluster using Hitachi Truecopy replication between two LDOMs running on pair of T4 servers ?
I'm thinking about 4-5 LDOMs on each T4 with Solaris Cluster frameworks installed in LDOM and connected in campus cluster configuration running NFS on ZFS formatted LUNs. To be able to failover I need HORCM to be able access command device on storage which should be raw LUN.
What actually concerns me is if it possible to pass raw LUNs to LDOMs necessary for HORCM operations ? We do not have T4 server with attached storage yet and I can not check if ldm add-io can operate on LUN level and not on HBA level and I yet to
Solaris Cluster in control domain and failover of LDOMs does not provide what I'm trying to achieve - HA for service running inside LDOM, whether it's NFS or Oracle.
Thanks !

Answer 1

Storm,
OK, I'm not sure I understood this statement:
"We can have same LUNs AA and BB zoned to node B and they also will be ReadWrite."
Suppose your zpool is called mypool. Further, suppose that that your LUN AA appears as /dev/did/dsk/d7 and LUN BB as /dev/did/dsk/d8. You would create the zpool as follows:
# zpool create mypool \
mirror /dev/did/dsk/d7 /dev/did/dsk/d8
You would then create a resource group and resource to control this.
# clrg create -n nodeA,nodeB my-storage-rg
# clrt register SUNW.HAStoragePlus
# clrs create -g my-storage-rg -t SUNW.HAStoragePlus \
-p zpools=mypool my-storage-hasp-rs
# clrg online -emM my-storage-rg
That would bring your zpool online on nodeA. The zpool would be read/write accessible from nodeA only. If you then switched the resource group to nodeB using:
# clrg switch -n nodeB my-storage-rg
Then you could read/write to the zpool from nodeB.
There is no way, either with mirrored zpools and HAStoragePlus or TrueCopy/UR & HORCM that you can read/write to a zpool from both nodes concurrently.
If someone forcible imported the zpool onto the other node, then you are very likely to suffer permanent zpool corruption. But then, if someone dd'ed over your root disk, that would be fatal too ;-)
With regards to replication via TC/UR and ZFS, the important feature for preserving data consistency is the UR journal. As far as I remember, these are only applicable to Universal Replicator when set up for asynchronous replication. I don't think UR supports journals with sync replication. However, since this isn't an Oracle product, I would strongly recommend you check my statements with Hitachi - things may have changes since I last spoke to them.
Hope that helps,
Tim
---