View 5.3
ESXi 5.5 - 1331820
vCSA 5.5.0.5201 Build 1476389
Non-persistent linked clones (refreshed on logoff)
I have 3 sites, all are setup identical with the following list above. The are all independent of each other so they are not sharing a vCenter or Composer.
I came in this morning and found that all 3 sites had desktops that were stuck at "Startup". I was able to ping, \\UNC to them, and able to view the desktops inside of the vCenter Console.
I had to recompose these desktops to get them up and running this morning. Currently on the phone with VMware support, but they aren't finding anything in the logs.
Anyone seen this before?
Yes, so here's what I found.
Our snapshot was taken before the DST change, and therefore it was an hour ahead of our current time. Since our linked clones get refreshed on logoff it reverts back to the snapshot with the old time. So when the linked clones get powered on they still have the old time, and the View Agent on the virtual desktop checks in with the View Connection Server with the old time (an hour ahead of our current time) saying it's starting up. Once it's up I'm assuming the Agent is telling the View Connection Server it is "Available", but the View Connection Server see's the new (correct) time which is an hour behind when it recorded the "Startup" time. Waiting an hour caused these "Startup" machines to show "Available".
I'm attaching a screenshot of the View Event Log that shows the Agent was starting up at 2:51pm (while it was really 1:51pm), and after waiting an hour the desktop changed from "Startup" to "Available".
I powered up my template last night, let the time update, and created a new snapshot. All of my desktops are working without issues this morning.
I would suspect DNS-problems, anything going at that time with network or name-servers?
// Linjo
Nope, no DNS issues. Each site has their own local domain controller too. Pretty weird for this to happen to all 3 sites.
Are the 3 sites totally decoupled? Nothing is connected like DNS, AD, IP-adressing, network?
If you have a desktop still in "startup" mode, check in the registery what connection-broker it got assigned and verify that it can reach that broker.
// Linjo
All 3 sites are connected to the same domain, which does hold DNS, but they all have local domain controllers.
I do not currently have a desktop still in startup mode.
VMware ticket #14451259303
But does the domain-controllers also serve as DNS-servers? (They are not by default)
If not then I would put some money on that the DNS what not available.
// Linjo
Yes, they are also DNS servers
This happened again to a desktop. It still shows Startup as the status. I was able to open a console session and logged in as local Administrator. I'm assuming the path you wanted me to check is HKEY_LOCAL_MACHINE\SOFTWARE\VMware, Inc.\VMware VDM\Agent\Configuration
It does have the brokers listed, and I am able to ping all of them.
How long have it been in that status?
Does it have any active connections to the brokers? You can check that with the "netstat -a" command.
Anything in the logs?
// Linjo
It's been at Startup for almost a couple of hours.
netstat -a shows it has established tcp/4001 to our connection server. We have 3 connection servers per site, does it matter which one is shows it's connecting to?
Nothing in the logs, I'm wondering if this a time issue since the time change. We've never had any issues in the past, but this is our first time change on View 5.3. VMware is recommending I take a new snapshot, and reboot the connection servers, composer servers, vCenter servers, and ESXi hosts.
This is indeed time related. I'll post more information tomorrow.
Hmm, so you think its related to the daylight saving time? Sounds reasonable if if started after Sunday.
Yes, so here's what I found.
Our snapshot was taken before the DST change, and therefore it was an hour ahead of our current time. Since our linked clones get refreshed on logoff it reverts back to the snapshot with the old time. So when the linked clones get powered on they still have the old time, and the View Agent on the virtual desktop checks in with the View Connection Server with the old time (an hour ahead of our current time) saying it's starting up. Once it's up I'm assuming the Agent is telling the View Connection Server it is "Available", but the View Connection Server see's the new (correct) time which is an hour behind when it recorded the "Startup" time. Waiting an hour caused these "Startup" machines to show "Available".
I'm attaching a screenshot of the View Event Log that shows the Agent was starting up at 2:51pm (while it was really 1:51pm), and after waiting an hour the desktop changed from "Startup" to "Available".
I powered up my template last night, let the time update, and created a new snapshot. All of my desktops are working without issues this morning.
Well done figuring that one out and thanks for the thorough conclusion.
Just to chime in we are also having this issue.