Performance reports

Discussion in 'Performance Questions' started by Shane B, Apr 20, 2016.

  1. Shane B

    Shane B Kilo Poster

    Messages:
    14
    Hello,

    Is there any way to easily pull performance metrics from the PCS hosts? Our customer is wanting to see this info. It would be helpful to actually be able to see this info at the host and guest level. Things like cpu usage over time, disk I/O, memory, etc.
     
  2. Pavel

    Pavel A.I. Auto-Responder Odin Team

    Messages:
    433
    Hello Shane,

    Virtuozzo itself does not store usage statistics. However, if you have the node registered in Virtual Automation you can easily find usage statistics over there. Just navigate to the VM overview screen, tap "Resources" button on the top panel, and you'll land on the page which will offer you some usage graphs.

    If you do not use Virtual Automation, you can start collecting stats on your own. Command "prlctl statistics <VMNAME>" would return you immediate usage for various counters.
     
  3. Shane B

    Shane B Kilo Poster

    Messages:
    14
    Here are results for one VM I checked. Is there a way to decipher what the numbers mean?

    guest.vcpu0.time: 13726194323830
    guest.vcpu1.time: 7679377618632
    guest.vcpu2.time: 10126139866551
    guest.vcpu3.time: 8788988644366
    guest.vcpu4.time: 23253878667957
    guest.vcpu5.time: 15044444448308
    guest.vcpu6.time: 40555722683883
    guest.vcpu7.time: 40687081858855
    guest.cpu.usage: 0
    guest.cpu.time: 419659
    host.cpu.time: 972275171
    ws.reclaimable: 1912
    guest.ram.usage: 2391
    guest.ram.cached: 80
    guest.ram.total: 8189
    guest.ram.balloon_actual: 0
    guest.ram.swap_in: 0
    guest.ram.swap_out: 0
    guest.ram.minor_fault: 17
    guest.ram.major_fault: 0
    devices.sata0.read_requests: 83347
    devices.sata0.write_requests: 1723654
    devices.sata0.read_total: 2585869312
    devices.sata0.write_total: 12080555008
    devices.sata1.read_requests: 3048
    devices.sata1.write_requests: 437387
    devices.sata1.read_total: 56865280
    devices.sata1.write_total: 3522447872
    net.classful.traffic 0 0 0 0 0
    net.classful.traffic 1 753356770 3809515 178491051 368300
    net.classful.traffic 2 0 0 0 0
    net.classful.traffic 3 0 0 0 0
    net.classful.traffic 4 0 0 0 0
    net.classful.traffic 5 0 0 0 0
    net.classful.traffic 6 0 0 0 0
    net.classful.traffic 7 0 0 0 0
    net.classful.traffic 8 0 0 0 0
    net.classful.traffic 9 0 0 0 0
    net.classful.traffic 10 0 0 0 0
    net.classful.traffic 11 0 0 0 0
    net.classful.traffic 12 0 0 0 0
    net.classful.traffic 13 0 0 0 0
    net.classful.traffic 14 0 0 0 0
    net.classful.traffic 15 0 0 0 0
    net.nic0.pkts_in: 155729934
    net.nic0.pkts_out: 471093
    net.nic0.bytes_in: 15991194526
    net.nic0.bytes_out: 187966429
    guest.fs0.total: 41401340
    guest.fs0.free: 19479876
    guest.fs0.name: Volume{01d60cef-cb78-4190-8a2b-c6f7a493d635}
    guest.fs0.disk.0: 0
    guest.fs1.total: 41809916
    guest.fs1.free: 26874328
    guest.fs1.name: Volume{2db096ff-aeaf-406d-9adc-da02b8a6fe63}
    guest.fs1.disk.0: 1
     
  4. Pavel

    Pavel A.I. Auto-Responder Odin Team

    Messages:
    433
    Hello,

    Most of these counters wouldn't be interesting for you, but you can find the ones you need, e.g.:
    guest.cpu.usage - % of host's CPU consumed by this VM. E.g. my host has 8 cores, and this VM, configured with 2 cores, runs a "stress -c 1" which occupies 1 core entirely. 1/8 =~12.5%. Note, percentage is relative to the node's power, not to VMs.
    net.classful.traffic X BI PI BO PO
    where X is the configured network class. If you do not have traffic shaping configured - all the traffic will be in "1".
    BI - Bytes In
    PI - Packets in
    BO - Bytes Out
    PO - Packets Out

    The rest is self-explanatory - ram, disk I\O, network I\O, filesystem's capacity
     
  5. Shane B

    Shane B Kilo Poster

    Messages:
    14
    Thanks for that. but the basic command does not specify any sort of time period. How do I know for what time period this information applies? For example, Customer A comes to us and says they think some VMs have been running slow and they would like to see performance metrics for the last 2 weeks. How would I grab this info and deliver it to them?
     
  6. Pavel

    Pavel A.I. Auto-Responder Odin Team

    Messages:
    433
    Dear Shane,

    As I stated initially, this command returns _immediate_ usage. If you want to get statistics you should call it, aggregate output, and store the data somewhere for further reference. Probably feed it to zabbix?
     

Share This Page