Lag inside of container

Discussion in 'General Questions' started by Shane B, Aug 10, 2016.

Tags:
  1. Shane B

    Shane B Kilo Poster

    Messages:
    11
    Hello,

    We have a customer who we moved from one datacenter with old hardware to a different datacenter with newer hardware and their performance decreased. This is 4.6 running on top of VMware ESXi 5.5. We thought it might be their connection latency, but we use Foglight in vWorkspace and don't really see any latency issues.
    Basically what they experience is input lag. They will type several words with nothing showing on the screen, and all at once 5 words will show up. Or they will be working in Excel and clicking between cells and the spreadsheet gets all messed up because it doesn't recognize the clicks correctly. The cpu usage of the host is fine, the storage performance is fine, the network is fine. Their average latency to the virtual desktop is 44 ms and their bandwidth averages about 8 Mb/s. My coworker accesses a virtual desktop on the same host and he swears he has no problem. I do see the same issues as the customer and I am in a different state then them.
     
  2. Pavel

    Pavel A.I. Auto-Responder Odin Team

    Messages:
    403
    Hello Shane,

    That's a complicated one... I'm pretty sure I faced something like that before, however, I cannot really recall what was the outcome.
    First of all - you've got to identify the scope. Is it only one virtual desktop that is affected, or is it true for every virtual desktop on that host? Try from both your and your coworkers computers. Also - do you use same OS as your co-worker? It could've been a matter of the RDP client version.
     
  3. Shane B

    Shane B Kilo Poster

    Messages:
    11
    There are 2 hosts in the cluster, and 2 Containers servers in the cluster. Originally both guest servers were on the same host, but we moved my guest server to the other host. There are only a small number of desktops running on each guest. Mine has the problem, and at least 2 users on the other Containers server are complaining of this. We all have the same symptoms. I use Windows 8.1 Pro, and they have used Win 7, 8.1, and 10 I believe. We have made sure they have the most recent version of the vWorkspace client, and I believe we have looked at the specific version of the RDP client in the past. I know we ruled that out in any case. I'm working with VCE to reseach an issue where 4 guests with Fault Tolerance enable have an extremely high "%VMWAIT" value. That is really the only metric I've seen that is outside normal operating parameters. My theory is that perhaps there is a scheduling problem on the physical cpus because of this value, but unfortunately nobody seems to really know the ramifications of the high %VMWAIT value.
     
  4. Pavel

    Pavel A.I. Auto-Responder Odin Team

    Messages:
    403
  5. Shane B

    Shane B Kilo Poster

    Messages:
    11
    HAHA, yeah I've studied that same article in depth, along with VMware's own documentation. The problem is that all other indicators are normal. There is no resource contention on cpu, memory, or storage. The single thing that can be pointed to is the very high %VMWAIT value for the guests with fault tolerance enabled. And the FT guests aren't the Containers servers. They just happen to be in the same cluster. It's perplexing, and of course the customer is not happy. Our other Containers servers on other hosts are perfectly fine. I'm just trying to get any input I can.
     
  6. Pavel

    Pavel A.I. Auto-Responder Odin Team

    Messages:
    403
    You should consider contacting support if that matter cannot really wait and causes frustration of end clients.
    Unfortunately problems like that are very difficult to investigate over-the-forum-conversation. Just like most of the performance-related issues.
     
  7. Shane B

    Shane B Kilo Poster

    Messages:
    11
    We've got one thing to try, and that is a small change on the hosts. (right now Fault Tolerance and vcpu hotadd are both enabled, and that is not a supported configuration. We've got to get vcpu hotadd disabled.)
    If that doesn't impact the problem then we'll open a ticket with Parallels/Odin support. Sorry, I have a hard time using the correct names for things
     
  8. Pavel

    Pavel A.I. Auto-Responder Odin Team

    Messages:
    403

Share This Page