1. Be sure to backup all data to be able to restore the system in the case something goes wrong. 2. Always do critical maintainance tasks like adding or removing volumes during offline hours.
Setting volumes offline
OCFS volumes that are in use can be set offline with ocfsutil. Setting a volume offline causes that the volume will be dismounted by the filesystem and that the “offline” state is written to the volume to prevent automatic (re-) mounting (which occurs during reboot and on several other events). Although the volume will be set offline on the node ocfsutil was executed the volume stays online on all the other nodes. This is true even if the volume already contains “offline” state. The reason is that there is no automatism that dismounts the volumes on the other nodes. Therefore one must take the volume offline on each node.
Setting volumes online
As discussed before setting a volume offline will prevent it from being mounted by OCFS because the “offline” state is stored on volume itself. One can initiate a mount by using ocfsutil to set a volume online. This just marks the volume with an “online” state and the filesystem driver will mount it when it receives the next mount request from the operating system (usually during boot, on first access or manually with the mountvol command).
Adding a new volume for use with OCFS
If the new volume is already available in the DiskAdministrator of all cluster nodes and the driveletter or link are already assigned consistently on all nodes one can just use ocfsformat to format the volume. Otherwise create the volume in DiskAdministrator and reboot all the nodes. After reboot check if the driveletter or link was assigned correctly and then format the volume using ocfsformat.
Removing a volume
Before removing a volume using DiskAdministrator a volume has to be set offline and deleted from OCFS usage.
1. Set the volume offline on all cluster nodes:
ocfsutil /c OfflineVol /m
2. Check if the volume is properly offlined by trying to access it. Do this on all nodes also.
3. Delete the volume from OCFS configuration:
ocfsutil /c DeleteVol /m
Upgrading OCFS
Upgrading OCFS is pretty easy. Basically one just has to copy new executables over the old and reboot the cluster nodes. Before upgrading it is important to check if the new OCFS driver is compatible with the current volume structure. Check all documents that come with the new release to ensure that. The general steps to upgrade the driver and tools are:
1. Rename the original OCFS driver on all nodes and backup the old tools e.g. ren c:\windows\system32\drivers\ocfs.sys ocfs.sys.old
2. Reboot all the nodes – OCFS will not be active after reboot
3. Install the new OCFS driver and the new OCFS tools on all nodes.
4. Reboot all the nodes
Follow these steps to guarantee no different OCFS drivers are used within the Cluster.
Note on OCFS 9.2.0.4 The shared disk access from one of the nodes is disabled suddenly which is a bug fixed in 9206 but in case you cannot upgrade to ocfs 9206 the solution is listed below
Let’s say our shared Disk has the label H:
Go to Disk Manager and try to locate H:If you see the H:\ but not his content on node1 (e.g. explorer asks to if you want to format it), then
- open a cmd box on node1- go to the %windir%\system32\osd9i\cfs - launch and collect the output of the commands:
ocfsutil /c dumpcfs -m H: /b test.txt
OcfsUtil /c OnlineVol /m H: /o
If you can't see the H:\ at all via Disk Manager or explorer, then it is a problem Concerning the OS and the vendor of the shared disks.
No comments:
Post a Comment