Query profiling on mysql cluster

Comments

3 comments

  • Avatar
    F Canepa

    Autoreplay...

    The problem was the theaded version ndbmtd which is definetely not an good option when using hyper-v virtual machine (even if each vm has 4 vcore). I don't know if iit's a good option when using vmware hypervisor.

    We switched to the single threaded version and now the queries whith distinct and group by are again very fast!

    The system was setup with the multithreaded version since day one... so I don't understand why the problem starts only after some months of activities

     

    0
    Comment actions Permalink
  • Avatar
    Johan

    Hi, 

    Thanks for letting us know. THere is probably due to contention. Also, is there anything else running on the physical host. Sometimes, if there are many VMs and many vCores sharing the same physical CPUs then you can get this behavior.

    How many cores you run on the physical hosts? How many vcores (summed up on all vms)  do you have on the physical host)?

    THanks

    Johan

    0
    Comment actions Permalink
  • Avatar
    F Canepa

    On the data layer vm we run only the ndb service but we don't know how many vm run on the same physical hypervisor. I suspect that running the threaded version of the ndb it's not a good option at all when using virtual machines.

    Thank you for your support

    0
    Comment actions Permalink

Please sign in to leave a comment.

Powered by Zendesk