3 hosts - 2 esx 4.0, 1 esxi 4.1, each host as 2 FC ports
2 FC san's - 1 netapp, 1 ibm ds3512, both are dual controller
2 FC switches, 1 link from each device to each switch
I created 2 arrays on the ds3512, then created 2 logical drives (lun0, lun4).
I added lun0 from the esxi 4.1 host and it shows up as a datastore on the other hosts.
If I add lun4 from the esxi 4.1 host, it does not show on either of the esx 4.0 hosts.
I am able to add the datastore from one of the esx 4.0 hsots, and it shows on itself and the esxi 4.1 host, but not the 2nd esx 4.0 host.
I've rescans hba's and refreshed.
Any thoughts? I can provide more detail if needed. Thanks.
I looked at the zoning on the FC switch. Each FC switch has one zone, and each FC port connected to a swtich is in the zone defined on that switch, including one FC link from each controller on the DS3512 to each switch (i.e., 4 FC links coming from the DS5312).
I wouldn't think that to be an issue or I wouldn't see one of the LUN's from the DS3512 on all of the hosts, but not the other LUN.
Also, the LUN should be presented as the same LUN on all hosts. When I created the datastore on the esx 4.0 host, it shows the same LUN as going through the wizard on the esxi 4.1 host.
One thing I do notice that may be related is the LUN that shows up on all 3 hosts is on Controller A in the DS3512, and the LUN that is giving me trouble is on Controller B in the DS3512. Again, though, I would think that would be an "all or none" issue with zoning and paths on the FC fabric.
Yes, the esx 4.0 host giving the most trouble sees LUN0 on one HBA and LUN4 on the other HBA, as I would expect, since they are on different paths and different controllers in the DS3512.
Again, if it were a path or LUN problem, I would expect the same result on all hosts....wouldn't I?
Stupid question here, but to the LUN's need to be unique across all SAN's in the FC fabric?
I have a LUN 0 and LUN 4 on my DS3512
I have 1, 2, 3, 10 from the NetApp.
(note: just be happen-stance, pure-dumb-luck, I happen to have unique LUN's)
On your ESX host, log into the console, and run esxcfg-volume -l.
If it comes back with anything (which should be the ds you don't see), run esxcfg-volume -M <VMFS label from previous command>
I may have discovered why the one host does not like the datastore, but I have no clue what to do from here.
From your post, I started looking at the datastore view and went to Manage Paths for each host.
It turns out that the 2 hosts (1 esx 4.0 and 1 esxi 4.1) have a path to controller B on the DS3512. The other host has an active path to controller A on the DS3512. The array and logical disk on the DS3512 are assigned to controller B. That explains (I think?) why the one host is having issues.
So if my hosts are
esx 4.0 #1 = host1
esx 4.0 #2 = host2
esxi 4.1 = host3
hba1 in each host connects to switch1
hba2 in each host connects to switch2
san-A1, san-B1 connect to switch1
san-A2, san-B2 connect to switch2
host1 --> hba1 --> switch1 --> san-B1
host2 --> hba1 --> switch1 --> san-A1
host3 --> hba2 --> switch2 --> san-B2
I hope that makes sense. The 2 working hosts each path through one of the two switches, but both path to controller B on the SAN. My problem host is pathing through switch1 and then to controller A on the SAN.
Hope that helps.
I can't easily reboot any of the hosts, and I'm not sure how I can reset one of the HBA's.
I have resolved this issue.
In the end, I upgraded the 2 esx 4.0 hosts to esxi 4.1 and the LUN's on the IBM SAN are all visible across all hosts.
That must be why ESX 4.0 is not certified for the DS3512.
Thanks all for the ideas and suggestions - great information to have!