There seems to be a problem with the Dynamic Discovery of one of my esx hosts. I'm not complaining, as this is fun stuff, but I've spent a week trying to configure a system that involves:
- one iscsi target server
- two esx 3.5 update 3 hosts
For my storage server, I have W2k8 running with VMware Workstation 6.05 containing a guest for Windows Storage Server with the iSCSI Target Software snap-in. Ok, so I have actually been doing this for a couple weeks now, and have tried working with guests running Starwind, and OpenFiler, but no avail.
Here's the strange part... each of these iSCSI target solutions I have been able to connect with from the second esx host in the two that I have. Though I have never been able to connect each esx host to the same iSCSI serving machine, even separate targets on the machine. My current topology is like this:
- Storage Server
- W2k8 (dataserver.web-wired.com)
- VMware Workstation 6.05
- Guest 1: Windows Storage Server (iscsi00.web-wired.com)
- IQN: iqn.1991-05.com.web-wired:isz-iscsi00-isz-esx01.web-wired.com-target
- IP: 192.168.xxx.173
- Guest 2: Windows Storage Server (iscsi02.web-wired.com)
- IQN: iqn.1991-05.com.web-wired:isz-iscsi02-isz-esx02.web-wired.com-target
- IP: 192.168.xxx.175 - Host Server
- ESX 3.5 U3 (esx01.web-wired.com)
- IQN: iqn.1998-01.com.vmware:isz-esx01-3c3d8db5
- IP (Console, vmswitch0): 192.168.xxx.180
- IP (VMKernel, vmswitch1): 192.168.xxx.172
- Guest 1: W2k8 (SQL Server)
- ESX 3.5 U3 (esx02.web-wired.com)
- IQN: iqn.1998-01.com.vmware:isz-esx02-702b1e5d
- IP (Console, vmswitch0): 192.168.xxx.187
- IP (VMKernel, vmswitch1): 192.168.xxx.181
- Guest 1: W2k8 (Domain Controller)
- Guest 2: CentOS 5.2 (SVN, Apache, Trac, WebDav)
I'm sure everyone loves the rescan button, especially after 3-5 minutes just to see Send Targets = 0. With the above configuration, esx02 is able to discover iSCSI targets on iscsi02, but esx01 is never able to discover targets on iscsi00, nor iscsi02 if configured to. I once had esx01 talking to StarWind targets (running on dataserver.web-wired.com), while esx02 was talking to iscsi02 (and iscsi01 which was running OpenFiler), but then I was not able to get esx02 to find targets on StarWind.
So to conclude, the reason I have two virtual machines each running a copy of Windows Storage Server is because I have yet to get two individual ESX hosts to find iSCSI targets running on the same storage server, OpenFiler, Windows, StarWind, etc. It occurred to me that perhaps Virtual Center can't do this, hence the two separate servers, but that seems ridiculous, as who would want to shell out $ for two storage servers when one should do? Anyway, it doesn't seem that is the real problem, more like the problem is that VC, or the ESX host has a problem finding targets.
To further, in each Wndows Storage Server instance, in the iSCSI Targer Software snap-in, on the tab that show iSCSI Initiators, I am able to see and add the IQN of each ESX host. Furthermore, within a separate Vista client, I am able to connect the iSCSI Initiator to the target of each iSCSI server. So it seems the iSCSI servers are doing their job.
If I had to guess, the ESX host is caching a bogus, or obsolete target, and any new changes aren't persisting. Please help?
I do have some linux knowledge, enough to poke around and do some damage anyway, so if there are some bash commands that might help me to diagnose, or config files I should be inspecting, please advise.
I am not sure what to tell you other than I have 2 esx hosts both accessing the same LUN from an Openfiler box. There are several sites out there telling you how to install openfiler and connect it via iscsi to esx. Then you just redo the steps for the other host.
Can you connect to a target from any of the VMs on ESX01 and ESX02?
Have you tried to ping and vmkping from the ESXes to the iSCSI targets?
Could just be a network problem.
I also use Openfiler with 2 ESX 3.5 hosts without a problem.
Ok, I tooled around with it some more, but no avail... so I decided I would just start it all over from scratch and reinstalled ESX on both machines, reinstalled the virtual center in a new vm, and reinstalled a clean OpenFiler. I prepped OF first, then went to my VI Client and created a new datacenter, and added the two hosts to it. Set up my vswitches and vmkernel to use the second physical nic, iSCSI firewall, discovery IP, etc, etc, and...
Both hosts, just like you both stated were able to connect to the same target and list the same LUNS. Not only that, but the rescan didn't take 5 minutes this time, more like 15 or 20 seconds.
Thanks for the inspiration, I think knowing it was possible reinvigorated me.