Jump to content


Photo

Error Starting VM, There were no servers available to complete the specified operation 5.6 FP1

Started by Mark Ward , 25 January 2011 - 07:16 PM
18 replies to this topic

Mark Ward Members

Mark Ward
  • 146 posts

Posted 25 January 2011 - 07:16 PM

Hello, I am in the process of upgrading 5.5 to 5.6 FP1. After reading all of the problems people have had doing an inplace upgrade starting with the Pool Master, I decided to try a different, what I thought would be safer, route. I currently have 3 hosts, 1 default pool, 1 iscsi storage device mounted. I moved all my VM's to the first two hosts, and then moved the last host out of the pool. All VM's are happy.

I then did an upgrade to the host I removed from the pool to 5.6 FP1 successfully. I have 6 nics. I bonded 0+1 and am using them to connect to a new iSCSI storage device. Nic 2 is my management nic as before the upgrade. I then created a new pool with my upgraded server. All 6 nics show up, the bond shows connected, the new storage is mounted, and all seems well. I decided to add a new VM to my host, server 2003 32 bit. I answer all of the questions in the wizard, and am using an iso bootable image in the cd drive of the host for the installation files.

When the new server shows up in the new pool, it is marked in Red and I get the following error.
*Error Starting VM, There were no servers available to start VM. Name host-c, Reason, Server cannot attach network (in the case of NIC bonding, this may be because attaching the network on this server would require other networks [that are currently active] to be taken down).*

I have tried breaking the bond and recreating, no difference. I have deleted the VM and tried a second time using a different nic. I am stumped. Any help would be appreciated. I did search the specific error, but the results seemed related to NIC's not showing up after the upgrade or not working. thank you.

Edited by: Mark Ward on Jan 26, 2011 4:41 PM



Luke Penrod Members

Luke Penrod
  • 75 posts

Posted 27 January 2011 - 04:36 PM

Mark,

Did you upgrade to 5.6, before 5.6 FP1 by chance? Is the pool master currently 5.5?

xe pool-dumpdatabase can backup your pool in case of a failure.

I would check out xe pif-list, xe vif-list, ifconfig. See if there are some clues of a hint in your /var/log/ directory.



Mark Ward Members

Mark Ward
  • 146 posts

Posted 27 January 2011 - 04:55 PM

Luke, thank you for taking the time to reply. I did not upgrade to 5.6 first. I went from 5.5.2 to 5.6FP1. The recommended way to upgrade scared me and I needed to add a new iSCSI storage device. My original pool had 3 hosts. I moved all of my running VM's off the 3rd host and removed it from the original pool. My reasoning was to upgrade the removed server to 5.6FP1, create a new pool, add the new storage device to the new pool, which worked fine and then start moving my VM's over time to the new pool. I intended to export them with zentools removed first and then import them into the new pool reinstall zentools and be good to go.

But before I started moving my production VM's I decided to try and install a VM from scratch as a test. So, to sum up, I have a new pool with one host running 5.6FP1, connected to the new iSCSI storage device without errors, two nics bonded and configured for the connection to the storage device and the 3rd nic of 6 total as the management connection. The error does not seem to reflect the problem completely. It seems it should be something simple. When the error reads, "There were no servers available to complete the specified operation."
What does that actually mean? I have a host/server, I have the start of the creation of a VM, and I have network storage communicating with the host. I have searched the internet, but not many solutions that are the same as my problem. I will look at the commands you suggested, but honestly, my experience with Xenserver is limited mostly to the Xen Center, not CLI.

If I can not determine the problem, I am inclined to do a clean install of XenServer 5.6FP1 rather than risk doing an inplace upgrade and having all of the other associated problems others have posted. It is a production environment. My main reason to upgrade is to eliminate the "Ghost" space that doing online backups is consuming. If not for that bug, I would be content to stay at 5.5. Any additional help would be appreciated. I noticed 35 views last time I checked, but you are the only one to venture a suggestion. Thanks again. mark



Luke Penrod Members

Luke Penrod
  • 75 posts

Posted 27 January 2011 - 05:08 PM

Mark,

Citrix recommends that you follow upgrades the following way:

"This means that to upgrade from version 5.0.0 to version 5.6, you first upgrade version 5.0.0
to version 5.5, and then upgrade to version 5.6, and so on."

Try removing the new Pool and add your adapters back and then try to create a new VM. I'm curious if something didn't take properly durning the upgrade path from 5.5 to 5.6 FP1 directly.

If you are failing I would then try a fresh install as you recommended. When you try to upgrade another server to join to the new pool, make sure you try the 5.5 to 5.6 to 5.6 FP1 path.



Mark Ward Members

Mark Ward
  • 146 posts

Posted 27 January 2011 - 05:20 PM

Luke, I recall reading the upgrade scenario you pointed out, but since I was at 5.5 with update 2, I assumed that 5.6FP1 would be a safe path to upgrade skipping 5.6 plain. During the upgrade, there was no indication that the install had a problem and I was easily able to attach the new network storage device. In the back of my mind, I can't help think that even though I have a new pool with an upgraded host, that some how the fact that it was in the old pool has left some low level trace that is causing this VM install to bomb out. I don't know enough about the software to really understand why this has happened.

The inplace upgrade on paper looks like a simple quick troublefree approach, but the majority of the recent threads seem to be related to failed inplace upgrades to 5.6. Storage not mounting, nic cards being offline etc. Most end up rolling back to an earlier version. My path/goal was to try and maintain my VM production environment regardless of how well the upgrade went. As it is, I have a problem, but my production enviroment is not experiencing any down time as a result of things not going completely as hoped/planned. That does take the pressure off to find a solution.Thank you again. mark



Luke Penrod Members

Luke Penrod
  • 75 posts

Posted 27 January 2011 - 05:29 PM

Mark,

Does it throw the same error if you try to create it on a local SR, instead of the iSCSI SR?



Mark Ward Members

Mark Ward
  • 146 posts

Posted 27 January 2011 - 05:58 PM

Luke, I am making some progress. I removed the pool. Still got the same error. I then broke the nic bond for my iSCSI traffice nic's 0 and 1 after removing my iSCSI interface and then readded the iSCSI interface using only one nic "0" and tried to run setup again. This time it worked. My original bond had the box unchecked to make available for VM's, but it seems there may be some conflict using a Bond with my first two nic cards. I am using the same configuration in my original pool where both hosts have the same two nic cards bonded for iSCSI traffic only. Odd. I am going to try and sneek back up on getting my VM installed recreating the pool first and then trying to recreate the Bond for ISCSI traffic. Thank you again. mark



Luke Penrod Members

Luke Penrod
  • 75 posts

Posted 27 January 2011 - 06:03 PM

Mark,

When you remove a server from a pool, it destroys the NIC ordering. Then it recreates it. I think your iSCSI SR on the bond may not be setup properly. If you disable the two ports on your switch temporarily, what happens? Can you still ping the iSCSI network from the xen command line? If so then that could be the issue and I can step you through recreating the NIC orders as needed.


Helpful Answer

Mark Ward Members

Mark Ward
  • 146 posts

Posted 27 January 2011 - 06:25 PM

Luke, unfortunately the HP switch I am using is web managed and I am not able to disable the individual ports at a quick glance. I understand what you are suggesting though I think. On the NIC tab of the management console, NIC 0 and NIC 1 are part of the motherboard, Broadcom brand and are physically obvious from the additional 4 port card I added with NIC 2 through NIC 5 and are the intel brand. When I create a new bond, I have done this a couple times so far, I am choosing to Bond NIC 0 and NIC 1. Once the bond is completed it is available to add as an interface for my iSCSI traffic.

So that I understand, you'd like me to disconnect my physical ports that I believe are carrying the iSCSI traffic and see if I can still ping the iSCSI storage device, not a loop back ping of the internal nics on my host? Any additional help would be appreciated. When I originally removed the host from my production pool, I don't believe I undid the Bond and after the upgrade the Bond was still in place. I may have removed it and readded it when I was connecting the new pool to the new storage device. Thank you again. mark



Mark Ward Members
  • #10

Mark Ward
  • 146 posts

Posted 27 January 2011 - 06:31 PM

Luke, one more thing. I am getting ready to add a new VM now that the new pool is created. I am in the networking section of the add VM wizard and have only 4 networks to choose from even without the bond. At this point shouldn't all 6 nics be available for choosing? I have attached to screen shots. mark

Attached Files



Luke Penrod Members
  • #11

Luke Penrod
  • 75 posts

Posted 27 January 2011 - 06:36 PM

I think it is grabbing the first four networks. Since, they only allow four during the initial creation. You could probably add the other two after you created the initial VM.


Helpful Answer

Luke Penrod Members
  • #12

Luke Penrod
  • 75 posts

Posted 27 January 2011 - 06:40 PM

You could do ifconfig eth0 down and ifconfig eth1 down from command line. Then try to ping the network. When your done just issue ifconfig eth0 up and ifconfig eth1 up.



Mark Ward Members
  • #13

Mark Ward
  • 146 posts

Posted 27 January 2011 - 06:51 PM

Luke, I have gone back to my production pool and tried to add a VM to see what Networks are offered. If you look at the attachment from my production pool wizard, you can see that Network 2 through 5 are available. Where on the new pool, I have Network 0 through 4. That must be the difference. On my production pool any VM's I add ignore the bonded network 0 and 1 and offer the available nics. I am going to try some of your suggestions, but on both pools the option to "add" network interfaces is greyed out. Hmmm?

Attached Files



Luke Penrod Members
  • #14

Luke Penrod
  • 75 posts

Posted 27 January 2011 - 06:58 PM

Mark,

In your first picture you posted it shows all the networks as "Auto = Yes." You need to have "Auto = No." if you click on the pool in XenCenter -> Network -> Properties on Network 0 and/or 4 -> Network Settings -> you will see a checkmark to automatically add to new VMs.

The reason it is greyed out is because you have to remove a network first.



Mark Ward Members
  • #15

Mark Ward
  • 146 posts

Posted 27 January 2011 - 07:18 PM

Luke, the solution was right in front of me. I did the eth up and down as individual nics and bonded doing the pinging with each time, but this time before I created the bond, I unchecked "available" for VM's and then created the bond. I now offered Network 2-5 same as my production pool. I am now installing a VM to my new host with my new storage and new pool. Thank you for all your help. I am going to try and give you credit for the fix. Thank you a ton. mark



Luke Penrod Members
  • #16

Luke Penrod
  • 75 posts

Posted 27 January 2011 - 07:30 PM

Glad I could help. Enjoy the new features.



Mark Ward Members
  • #17

Mark Ward
  • 146 posts

Posted 27 January 2011 - 07:54 PM

Luke, I often learn more when things don't go right the first time. The hard work is ahead, but I am looking forward to the new features and improvements. Thank you again. mark



Harrison hui Members
  • #18

Harrison hui
  • 1 posts

Posted 12 February 2011 - 02:12 AM

I met a same question.
I have install xenserver 5.6FTP1, and setup BIOS CPU's VT(Virtualization Technology) enable.
but, when start VM report error.
Could you help me!
Thank you very much!

Attached Files



Luke Penrod Members
  • #19

Luke Penrod
  • 75 posts

Posted 16 February 2011 - 05:42 AM

Harrison,

What processor and motherboard are you using?



Give Us feedback