This discussion has been locked. The information referenced herein may be inaccurate due to age, software updates, or external references.
You can no longer post new replies to this discussion. If you have a similar question you can start a new discussion in this forum.

VMware – managing virtual switches inside your ESX Server

FormerMember
FormerMember
From talking to customers, we know that many-possibly most-of you work in organizations that have embraced virtualization to some extent.  VMware's ESX server is especially popular.  One interesting aspect of ESX servers that hasn't gotten a lot of attention are the virtual switches inside that ESX server.  We here at SolarWinds are wondering what the Thwack community thinks about virtual switches and virtual switch management.
  • As network engineers, how do you manage those virtual switches?
  • As network engineers, do you monitor virtual switches?
  • What visibility do you have into the network traffic inside the ESX server?
    • Do you wish you had more?
    • Have you ever experienced problems related to traffic inside the ESX servers?

Virtualization is only going to grow, and we'd like to hear your thoughts on how its impacting networking and network engineers.

  • Denny-

    It's a good thought, but from what I understand, the VM switches are generally dumb. They aren't configurable (like a physical switch), aren't manageable, they don't pass STP, etc. I have never experienced an issue with the virtual switches, but our VM guy is pretty good so he knows how to configure them. Even on my test system, I've never had a problem, but I have not purposely tried to break it either. I'd be interested in hearing if/how we can monitor them.
     
    Collin 

  • We currently are using VMware's ESX Servers.  We are not monitoring the virtual switches.  Should we?
  • FormerMember
    0 FormerMember in reply to ceclark
    the VM switches are generally dumb. They aren't configurable (like a physical switch), aren't manageable, they don't pass STP, etc



    True, but don't worry about the how.  

     

    I have never experienced an issue with the virtual switches


    That's a good data point.

    but I have not purposely tried to break it either


    Do you usually try to break your stuff?  emoticons_wink.png

    I'd be interested in hearing if/how we can monitor them.
     

    What would you want to know?  What would make monitoring it a priority?

  • FormerMember
    0 FormerMember in reply to JeanC.
    We are not monitoring the virtual switches.  Should we?
     

    I don't know.  Should you?

  • but I have not purposely tried to break it either


    Do you usually try to break your stuff?  ;-)

     Try sometimes, generally break, yes!

    I would not see it as a priority for me. I would rather see enhancements in other areas (integration with Cirrus, web admin on Orion, etc)

  •  I personally would like to have a better handle on the traffic that flows across the vSwitches in our VMware environment. We have a number of servers that our application staff swears is our fault for performance. We are looking at additional tools to dive into the server and disk systems in VMware, better network information would be beneficial as well.

     As an aside, if the APM also provided information about the VMware environment, that would be beneficial. Again, our application developers have figured out how to write code in such a way that 100% of the fastest machines on the planet are still not enough to handle their code.

     Details include, vCPU time assigned, Real CPU time, I/O wait time, Memory stats such shared pages, the ballon. On top of all this the reservation vs available statistics.

    I would really need to sit down and think about what stats would be the most important. I bet if we had them all I could find a nugget of useful information like I did with Orion and Netflow when I discovered the 2 to 3GB file copies across a T-1 every time a person logged on.

  • We have several VM ware server here with multiple VM switches.  We would like to see up/down, traffic flow, packet loss to them.


     


    Yea I can see the guest servers and their nic's but seeing this info for the host would be great.

  • Physical servers with dedicated networks ports are rapidly becoming a thing of the past. On a daily basis our organization is virtualizing its physical servers. This is making network services job of providing feedback/performance monitoring very difficult. Any and all visibility, performance reporting that Orion can do for ESX infrastructure is so desperately needed.

  • I would like to see at least the same information for virtual switch interfaces that we can currently see with physical switch interfaces in Orion. Being able to identify the VMs that are the top bandwidth users would assist me in telling the Server Admins that the network is not the problem. We haven't experienced any problems yet, but I just know (based on history) that it is only a matter of time.

  • Indeed, I agree, I think it would be phenomenal to be able to get statistics from the VM Switches.

    Then I'm guessing that VMWare would have to create some form of management access to them.

    To answer your questions:

    1) The server teams manage the VM Switches.
    2) Network teams don't monitor them.
    3) No visibility.
    4) Yes wish Network Teams were able to look deeper into them.
    5) Yes.. Had to monitor at the aggregate port on the Cisco switch that the server was connected to.

    I personally think we need to see the VM Switches managed like any other switch on the network, with relevant security controls, port mirroring etc. that could be enabled.