Hello, I need to test whether our product will work with GPFS filesystems and I have some questions regarding the setup:
Do I need to dedicate an entire hard disk if I want to have GPFS on it? Or can I somehow split a disk into 2 virtual disks, and only use 1 for gpfs?
If lspv returns "none" that means the disk is connected but has not been allocated, and thus available for use in GPFS setup?
Example (hdisk3):
[root@aix01: /]# lspv
hdisk0 00ca0c3546524b4e rootvg active
hdisk1 00ca0c35797dab06 vg1 active
hdisk2 00ca0c35797daba9 vg1 active
hdisk3 none None
hdisk4 00ca0c35645c51bf san1 active
hdisk5 00ca0c3564646ef3 san2 active
Do I have to have a second node in order to create a cluster and create a gpfs on that cluster? Or can I start with one node (just to test functionality) and add a second node later? (basically, can I use that one system above with hdisk3 available to test gpfs on its own?)
To your first question, GPFS is a full device allocation, therefore, anything you wish to allocate will be consumed as an NSD.
Regarding your second question, yes, that disk is more than likely empty and therefore can be used for GPFS.
Finally, GPFS can run on one node (a single node quorum) and can have additional nodes added to it later. Make sure you have good time synchronization (xntpd) and decent performing disk storage.
Best regards,
jstnobdy
---------- Post updated at 09:02 PM ---------- Previous update was at 08:57 PM ----------
One more thing, GPFS is pretty tolerant, just make sure you define "dataOnly" and "metadataOnly" disks when you create your NSDs (mmcrnsd). In this case, allocate 'x'-amount of storage for dataOnly NSDs and a relatively small disk for metadataOnly.
Good luck. If you have any questions, let me know.
