How to determine which Client is busy?

Discussion in 'Post-Deployment: Configuration and Setup' started by Machiasiaweb, Jun 23, 2016.

  1. Machiasiaweb

    Machiasiaweb Mega Poster

    Messages:
    178
    Hello,

    While checking pstorage top

    It can showing top busy Client ID like below

    ===
    CLID LEASES READ WRITE RD_OPS WR_OPS FSYNCS IOLAT(ms) HOST
    14436 4/94 85KB/s 7.72MB/s 4ops/s 310ops/s 39ops/s 1.52/249 pcs1
    26889 0/27 845KB/s 2.37MB/s 14ops/s 60ops/s 12ops/s 1.45/402 pcs2
    15259 5/138 290KB/s 2.21MB/s 5ops/s 167ops/s 51ops/s 1.08/250 pcs2
    ====
    but how can I determine which CT or VM are busy?

    Seems manual did not mentioned
    http://download.parallels.com/doc/pcs/html/Parallels_Cloud_Storage_Administrators_Guide/35918.htm

    Thanks!
     
  2. Pavel

    Pavel A.I. Auto-Responder Staff Member

    Messages:
    475
    Hello,

    You can monitor load on the pstorage mount point using following command:
    Code:
    # pstorage -c %clustername% mnt-top
    Run it from the node where you want to investigate the load
     
  3. Machiasiaweb

    Machiasiaweb Mega Poster

    Messages:
    178
    Hello,

    Could you please explain more about it. Like following example output of mnt-top

    w 1 3ops/s 71KB/s 16ops/s 1.71MB/s 1ops/s 0 ...t.hdd/root.hds
    w 1 11ops/s 759KB/s 0ops/s 0B/s 0ops/s 0 ...t.hdd/root.hds
    w 1 1ops/s 3KB/s 66ops/s 400KB/s 1ops/s 0 ...t.hdd/root.hds
    w 1 1ops/s 3KB/s 13ops/s 269KB/s 6ops/s 0 ...-4b5240492d11}
    w 1 0ops/s 0B/s 13ops/s 214KB/s 4ops/s 0 ...t.hdd/root.hds
    w 1 0ops/s 0B/s 30ops/s 208KB/s 1ops/s 0 ...t.hdd/root.hds
    w 1 0ops/s 0B/s 6ops/s 138KB/s 2ops/s 0 ...t.hdd/root.hds
    w 1 0ops/s 0B/s 2ops/s 129KB/s 0ops/s 0 ...t.hdd/root.hds
    ... (rows 1-8 of 142)
    BOOT_BLKS MAIN_BLKS HITS_LAST HITS_AVER AGE PATH
    1 0 100.0% 100.0% 16d 20h /private/1462/.owner
    1 0 100.0% 100.0% 16d 20h /private/2361/.owner
    1 0 100.0% 100.0% 16d 20h /private/2521/.owner
    1 0 100.0% 100.0% 16d 20h /private/3281/.owner
    1 0 100.0% 100.0% 16d 20h /private/2445/.owner
    1 0 100.0% 100.0% 5h 48m .../templates/webalizer/timestamp
    1 0 100.0% 100.0% 5h 48m .../root.hdd/templates/vzpackages
    1 0 100.0% 100.0% 5h 48m ...mplates/pp-webmail/timestamp
    ... (rows 1-8 of 728)
    CSID STATUS RDOPS WROPS SYNC NETLAT IOLAT
    1065 connected 0ops/s 5ops/s 4ops/s 0.14/0 5.25/32
    1063 connected 0ops/s 4ops/s 3ops/s 0.15/0 8.37/46
    1042 connected 0ops/s 7ops/s 3ops/s 0.15/0 0.18/10
    1057 connected 0ops/s 7ops/s 3ops/s 0.14/0 0.16/0
    1068 connected 0ops/s 1ops/s 2ops/s 0.17/0 4.23/54
    1055 connected 0ops/s 13ops/s 9ops/s 0.15/0 0.18/0
    1052 connected 0ops/s 20ops/s 2ops/s 0.16/0 0.17/0
    1040 connected 0ops/s 28ops/s 4ops/s 0.16/0 0.36/12

    -----------------
    thanks!
     
  4. Pavel

    Pavel A.I. Auto-Responder Staff Member

    Messages:
    475
    mnt-top shows 3 tables to you:
    1) fstat table only
    2) cache hit table
    3) chunk servers table

    We're only interested in the first one - this table shows read/write stats for the files from your mount point.
    Press "f" to show just the first table.
    Make sure your terminal is wide enough to fit in the data - your output is quite truncated.
    For example, here's output from my cluster (I've got only 1 VM running, so you wont see much over there):
    Code:
    RW HNDL      RDOPS    RDBYTES      WROPS    WRBYTES       SYNC PENDREQ PATH
    w    2     0ops/s       0B/s     1ops/s      2KB/s     0ops/s       0 /vmprivate/pvo-w2k12r2.pvm/pvo-w2k12r2-0.hdd/pvo-w2k12r2-0.hdd.0.{5fbaabe3-6958-40ff-92a7-860e329aab41}.hds
    w    1     0ops/s       0B/s     1ops/s     360B/s     0ops/s       0 /vmprivate/pvo-w2k12r2.pvm/parallels.log
    
    You can see all the stats per file. Look at the files path - these files belong to pvo-w2k12r2 virtual machine - you can see it by looking at the path, thus I know that this I\O is caused by this VM.

    It's the most reliable way to check I\O for the mount point. Of course you may look at "pstat" and monitor disk I\O there, however, it will only show VM's activity - if I\O is caused by something else, not by the VM itself, you won't see it there.
     

Share This Page