VMware vSphere

 View Only
  • 1.  VM freezes (stun) at *creation* of snapshot

    Posted Mar 12, 2015 10:22 AM

    VM's  running on several ESXi 5.5 Update 2 hosts freeze at the *creation* of a snapshot. The freeze periode depends on the VM. Some VMs lose 3 pings.

    Some VMs even have a freeze period of 1 minute! Network connection is lost during freeze.  Internal memory is not included in the snapshot.

    We even observe BSOD on some Windows Server VMs (particular those with Exchange Server installed)

    When snapshots are deleted we observe same freeze period even when there is no IO activity during the snapshot is active. A VM with 16GB of internal memory can be frozen for over 4 minutes.

    Appreciate any hints in troubleshooting this issue



  • 2.  RE: VM freezes (stun) at *creation* of snapshot

    Posted Mar 12, 2015 11:22 AM

    Are you deselecting the "Snapshot the virtual machine's memory" option when creating the snapshots?  That improved things for us.  We have virtualized Exchange too, but the stunning only used to occur when comitting the snapshots.



  • 3.  RE: VM freezes (stun) at *creation* of snapshot

    Posted Mar 12, 2015 11:28 AM

    Yes when even deselecting the "Snapshot the virtual machine's memory" VMs are stunned during the creation of the snapshot.



  • 4.  RE: VM freezes (stun) at *creation* of snapshot

    Posted Mar 12, 2015 11:35 AM

    Okay, how are you presenting storage to the host(s)?  It almost sounds like a locking issue with the datastore.



  • 5.  RE: VM freezes (stun) at *creation* of snapshot

    Posted Mar 12, 2015 11:37 AM

    Hi,

    Need to understand whether you have sufficient resouces available into your environment like CPU, RAM and Storage end. Usually that impact this type of issues.



  • 6.  RE: VM freezes (stun) at *creation* of snapshot

    Posted Mar 12, 2015 11:44 AM

    Also, had you only started to utilize snapshots and experienced the stunning or had you been using them with no issues until now?  If things just started going south, what changed?  Have you overcommitted resources on the host or datastore?  Is it only occuring on a single datastore and could there be too many hosts accessing that datastore concurrently?  Lastly, does the snapshot process complete?  Do you see the -000001.vmdk file generated in the VM's directory?