Hi,
How long should it take to create a datastore? Or how can one tell that something is not quite right?
I have tested and tabulated the results, and only 1 ESX server is always the fastest. Whether it's creating or deleting a datastore from the same LUN, it will take an average of 3min.
The rest of the ESX servers will take anywhere between 14-18 min to create and 10-12min to delete.
Btw, this recently happened after upgrading to ESX 4.1 Upd 1.
I did a fresh installation on Hosts B3 & B4, but ONLY Host B3 gave the "3min" average.
Any fresh ideas are welcomed.
Here's the setup & test results:
ESX 4.1 Upd 1 b348481
SAN: EMC NS-480 (tried on NetApp as well. same result. so no problem with the SAN & SAN Ports)
LUN: 144GB FC
Datastore block size: 4 (didn't matter, tried with other block size on the slowest host & still the same results)
MPP: MRU
Storage Array Type; VMW_SATP_CX (default)
Cluster: A -> 2 hosts
Host A1: HP ProLiant DL380 G6 / HBA - QLogic 2432
Host A2: HP ProLiant DL380 G6 / HBA - QLogic 2432
Cluster: B -> 4 hosts
Host B1: DELL PE R900 / HBA - Emulex LPe11000
Host B2: DELL PE R900 / HBA - Emulex LPe11000
Host B3: DELL PE R910 / HBA - QLogic 2432
Host B4: DELL PE R910 / HBA - QLogic 2432
Cluster: C -> 2 hosts
Host C1: HP ProLiant DL580 G5 / HBA - QLogic 2432
Host C2: HP ProLiant DL580 G5 / HBA - QLogic 2432
Seems you have some locking issue.
But strange that one host works fine.
All your hosts see the same LUNs?
How much LUNs, hosts and VMs do you have?
Andre
Hi Andre,
As per the earlier post, there are 8 ESX hosts in 3 different clusters.
All of the hosts sees the same LUNs and there a total of 150++ VMs spread across the 8 hosts.
I have NetApp & EMC SAN as datastores.
This series of tests were done on the 144GB FC LUN on NS-480(EMC) for creating/deleting datastore. No VMs are on that LUN.
When you mentioned "locking issues", are you referring to the host locking on LUNs? I didn't see any "SCSI Reservation" sort of errors logged in the vmkernel. I did restart the mgmt-vmware services on all hosts and the vCenter svcs as well. But none seemed to have an effect on the problem.