UNSOLVED
Help! MD3000 + 2x PE 2900 problem with W2k3 MSCS
I attempt to implement a two-host application with Windows 2003 Ent SP2 Cluster Service
The configuration is shown below:
I hope Host 2 would take over Host 1's functionality during Host 1's failure and vice versa
I've finished installing windows 2003 and cluster service on both nodes and configured the RAID in MSDM. Everything seems OK. However, I cannot move group to another node in Cluster Administrative. I check the event viewer, the error messages like
'The disk associated with cluster disk resource 'Disk F:' could not be found. The expected signature of the disk was ....' (e.g, I hope move a group with Disk F)
I found that if I set the preferred path of Disk F to host 1 In MSDM, I can immediately see the Disk F in My Computer of host 1, but cannot move group with Disk F to Host 2.
Also, if I set the preferred path of Disk F to host 2, I can see Disk F in My Computer of host 2, but cannot move group to host 1.
Why the host connected to the non-preferred controller cannot take control of the MD3000 Disk?
Many Thanks!
343
0
Posted August 7th, 2009 10:00
it' like, if I set the preferred path to Controller 0 (connected to Host 1), Host 1 would see the disks any way. But in Host 2, there's no disks appearing in My Computer, and I can see them it the Disk Management (diskmgmt.msc) with 'Not initialized' sign. I would fail if I try to initialize them.
So far, the only way to make Host 2 to see the virtual disks in My Computer is to set the preferred path to Controller 1 (which is connected to Host 2). i.e, the disk groups cannot change their ownership to the non-preferred host AUTOMATICALLY
I do make the hostgroup in MDSM and map all virtual disks to the group, not indivial server.
BTW, I can see a 20M disk in Disk Management on both hosts. I probably know it's some kind of management disk for MD3000. I'm not sure if i have done something on it like initiailzation or convert it. Is it possible that i damaged this disk?
343
0
Posted August 7th, 2009 11:00
Being able to see the 20M disk (Access Lun) and the fact that the LUNs are not being moved between controllers indicates that you do not have the MDSM software installed on the cluster nodes.
You should take the following steps:
1) In MDSM check which version of the NVSRAM is running on the array. You will want to make sure it is the versioon that ends with 7xx. The 0xx version is used when any host has more then one cable.
2) On one node install MDSM, either full or just the host. Then reboot.
3) Repeat step 2 on the other node.After this the 20MB luns should only show up in device manager under disk drives as the Universal X-Port and failover of the cluster should now work.
-cjtompsett
343
0
Posted August 7th, 2009 11:00
Thx!
I'll try you solution next Monday~ that sounds reasonable and professiona:emotion-21:l
Dev Mgr
6 Operator
•
9308 Posts
•
3 Points
343
0
Posted August 7th, 2009 08:00
Are the disks visible (not necessarily readable) on both servers?
In other words; on one node you should see all drives with their drive letters and such. On the other node you should see the disks and partitions, but it'll say 'unreadable' or something like that. Verify that both servers can see the same drives.
Did you make a hostgroup in the MD3000 management application (MDSM) and put both servers in that host group and assign all virtual disks to the group (not an individual server)? Also, go to the modify tab, select to modify the host topology and then pick each server and verify the servers are set for Windows clustering.