Contact Us

support@onapp.com

U.S: (+1) 888-876-8666

UK: +44 (0) 203-318-5364

Need the ability to customize kvm guest domain config

Completed

Comments

3 comments

  • Official comment

    Hello

    You can manage your virtual servers at a raw configuration level by editing the XML configuration file available for each VS in your Control Panel. The functionality is available for virtual servers that are built on Xen and KVM compute resources. 

    https://docs.onapp.com/agm/latest/appliances/virtual-servers/manage-virtual-server-advanced-configuration

     
    Comment actions Permalink
  • Avatar
    Kevin Lee

    The type of customization needed here is pretty standard across multiple virtualization platforms, and part of the standard API function of both KVM and XEN. The changes needed here are to correct for an almost 500% performance degradation we are seeing, so this is a serious impact, that should NOT be ignored. This should be something submitted as a BUG rather than a feature request since the static settings are causing problems. The settings that should be for each KVM domain we spin up should be the following for the HARD DISKS

    <disk type='block' device='disk'>
    <driver name='qemu' type='raw' cache='writeback' io='native'/>
    <source dev='/dev/onapp-oxbh403bh9q0/a24livnnc10vrm'/>
    <target dev='vda' bus='ide'/>
    <address type='drive' controller='0' bus='0' target='0' unit='0'/>
    </disk>

    The settings that need to be added are as follows;
    <driver name='qemu' type='raw' cache='writeback' io='native'/>

    This setting right here allows the I/O to not have to navigate through the Virtualization software driver and instead. This setting is detailed in several RedHat documents

    http://www.redhat.com/summit/2011/presentations/summit/decoding_the_code/wednesday/wagner_w_420_kvm_performance_improvements_and_optimizations.pdf

    This should offer all of the evidence that these static settings that you don't allow to be user modified or template based, are more of a bug since this type of tuning is different for each environment and hardware, and the settings that we are left with give us a bottleneck of no more than 50-60MB/s I/O on every KVM VM.

    Please get a higher level engineer involved here so that this can be properly reviewed and a resolution found. 

    0
    Comment actions Permalink
  • Avatar
    support58

    Not having the option to tune KVM block device configuration is absurd. Depending on the backing(storage+disk type) the configuration for best performance is very different. The default values of the current setup(not having io option to native or threads) is a HUGE bottleneck and must be addressed now.

    0
    Comment actions Permalink

Post is closed for comments.