I've recently installed a vSphere 4.0 setup connected to a cx3-20 with flare 126.96.36.199.019. As per the Host Connectivity guide, I did not install the naviagent on the ESX consoles. I've got two problems that I'm running into
Any help would be appreciated.
I have IP connectivity. Previously with navisphere, I had to open up ports in the ESX firewall, however, I haven't seen any instructions that require that this be done with vsphere.
Once a host is manually registered to change back you need to shut down the host, remove the host from the Storage Group, de-register the host in Connectivity Status (all paths listed), then turn on the host.
The issue with NaviAgent use to be finding the correct NIC to use to communicate with the array, EMC agent uses the first bound NIC, so for example, on a Windows host if you have 4 NICs and the one that you use to communicate with the array is the second NIC bound, Agent will push to the array the first NIC bound and Agent on the array will attempt to talk to the host using the wrong NIC. It might be the same for the agent in V4.
So I followed your procedure. I shut down the vsphere system, removed the host from all storage groups, went into Connectivity status and deregistered the host. Then I powered the host back on. The host has only one nic in it that is connected to the network and configured with an IP address. When the host came back on, I received the expected event that two unregistered hbas logged into the fabric. However, when I went to the hosts tab, I saw once again the host and it was listed as Manually registered.
Is it also expected that you cannot do things like "Server Update now" since there's no agent on the vsphere?
The Agent on the host that's included in V4 is suppose to act the same way that the Navisphere host Agent works - the array contacts the host agent using TCP port 6389 (check that these ports are open and no firewall), and the host agent is suppise to respond.
You may need to open a case with VMware to see what's going on if the TCP port is open and still not working.
I have read that the vsphere registration is a "push" from the ESX kernel to the Clariion. I can not find any documentation indicating TCP port 6389 must be open for the ESX 4.0 registration process, only for Navisphere agent(which we do not install on ESX4.0 per EMC recommendation), and or Navisphere CLI which at this point we do not install on ESX 4.0.
We are having issue with the auto-registratioin of ESX4.0 to our CX4-480 array's(Flare v29). We did open port 6389 on one ESX 4.0 host to asist in troubleshooing the issue. Still no auto-registration. We have also verified within the ESX configuration/advanced settings/disk, the Disk.EnableNaviReg setting was set to "1" to allow the auto registration.
It seems we are missing some small piece to this puzzle, any ideas would be appreciated.
Ok.. so I hooked up a brand new server today with HBAs that have never been seen by the clarrion and made an interesting observation.
The new host (2 hbas) automatically appeared as "manually registered".
The other hosts which were upgraded from ESX3.5 to vsphere4, which previously had automatic registration via the agent, all show "U" icons for agent not reachable. Seems that the clariion thinks that an agent should be there where there isn't an agent anymore and that the clarrion can't switch from agent to no agent when you upgrade from esx3.5 to 4.
I've been looking at this issue on and off for a while and I found a Primus article from November 2009 (can't remember the ID - sorry) stating that with vSphere 4, the kernel will auto-register with the Clariion array but that on FLARE revisions 28 or less, the host will display "Manually registered; Host agent not reachable" in the Navisphere storage management GUI - on FLARE 29 it should read as fully registered.
Take a look at EMC Knowledgebase solution emc228353, this related to release28 but may also be needed for release29. If this is the case, I would suggest you notify EMC via a service request that the article needs to be updated for R29 if still affected or an alternate registration process is needed.