AIX

AIX

Connect with fellow AIX users and experts to gain knowledge, share insights, and solve problems.


#Power
 View Only
  • 1.  HBA adapter I/O reporting by nmon

    Posted Tue April 04, 2006 09:52 AM

    Originally posted by: SystemAdmin


    I noticed something very peculiar with regards to the interactive output from nmon on the [b]fcs0,fcs1[/b] adapter stats. For some reason even though the HBA cards are doing balanced I/O (separate attachment from switch stats), [b]fcs1[/b] seems to be display the cumulative I/O and [b]fcs0[/b] with none of the I/O. I don't recall seeing this in versions previous to 10 (it's happening in [b]10r[/b] as well as [b]11d[/b]).

    We are running AIX 5.3 ML3 (64 bit) as well as the same model of HBA's. Please let me know if you need any more information regarding our configuration and/or whether anyone can confirm this.

    code


    N N M M OOOO N N For online help type: h
    NN N MM MM O O NN N For command line option help:
    N N N M MM M O O N N N quick-hint nmon -?
    N N N M M O O N N N full-details nmon -h
    N NN M M O O N NN To start nmon the same way every time?
    N N M M OOOO N N set NMON ksh variable, for example:

    export NMON=cmt
    Version v11d for AIX53
    8 - CPUs currently
    8 - CPUs configured
    1902 - MHz CPU clock rate
    PowerPC_POWER5 - Processor
    64 bit - Hardware
    64 bit - Kernel
    Dynamic - Logical Partition
    5.3.0.40 ML03 - AIX Kernel Version
    cernprd3 - Hostname

    nmonp=PartitionsHost=cernprd3Refresh=4 secs17:23.12
    Adapter-I/O
    Name %busy read write xfers Disks Adapter-Type
    fscsi0 8.0 0.0 295.7 KB/s 55.9 149 FC SCSI I/O Controlle
    fscsi1 8.0 0.0 255.7 KB/s 55.9 149 FC SCSI I/O Controlle
    ide0 0.0 0.0 0.0 KB/s 0.0 1 ATA/IDE Controller De
    fcs1 8.0 0.0 543.4 KB/s 111.9 147 FC Adapter <-- this looks to be total between
    fcs0 0.0 0.0 0.0 KB/s 0.0 1 FC Adapter fcs0 and fcs1
    sisscsia2 7.0 0.0 47.9 KB/s 12.0 2 PCI-X Ultra320 SCSI A
    sisscsia4 0.0 0.0 0.0 KB/s 0.0 1 PCI-X Ultra320 SCSI A
    TOTALS 7 adapters 0.0 1142.7 KB/s 235.7 450 TOTAL(MB/s)=1.1
    [/code]
    #AIX-Forum


  • 2.  nmon questions are better placed on the performance tools forum

    Posted Thu April 06, 2006 05:00 AM

    Originally posted by: nagger


    nmon is just reporting what is returned from the AIX libperfstat library. In nmon 9 and earlier attempted to work out the connections between adapters and disks the hard way. In nmon 10 onwards this function is provided by libperfstat. So this might be a case of GIGO.

    There is a simple tool with source code that can be used to report the results returned from libperfstat. You can download it from http://www.ibm.com/collaboration/wiki/display/WikiPtype/ryo
    called disks.

    If this reports the same confirguration as nmon then feel free to report this to AIX Support as a PMR to be fixed. I think the problem is that the disks are reported as they were initially discovered when AIX started rather than as they are actually working. I think there is a way to balance them out but I do not know the details.

    In the mean time, you can use nmon's User Defined Disk Groups to report the disks in anyway you like - see the nmon manual for details.
    #AIX-Forum


  • 3.  Re: nmon questions are better placed on the performance tools forum

    Posted Fri April 07, 2006 12:15 PM

    Originally posted by: SystemAdmin


    So after running the [b]adapt[/b] utility, I see the following -

    code
    Statistics for adapter : fcs1 which is adapter number=4
    description : FC Adapter
    number of devices : 147
    Size and Free (MB) : 2043392 269936
    Capability KB/sec : 0
    transfers : 647952846 delta=22
    [b]number of blocks read : 131520702141 delta=18
    number of blocks written : 1812357700 delta=409[/b]
    time : 505477001 delta=0
    Statistics for adapter : fcs0 which is adapter number=5
    description : FC Adapter
    number of devices : 1
    Size and Free (MB) : 11488 320
    Capability KB/sec : 0
    transfers : 4084358 delta=0
    [b]number of blocks read : 1002834410 delta=0
    number of blocks written : 139792 delta=0[/b]
    time : 4891843 delta=0
    [/code]

    Would this confirm that the library routine is incorrect?
    #AIX-Forum


  • 4.  Oh it is correct

    Posted Tue April 11, 2006 03:51 AM

    Originally posted by: nagger


    Well, the numbers are correct as far as AIX is concerned.

    As you can see 147 devices on fcs1 and 1 on fcs0 are being reported.
    The problem is, this is not what you might have expected.

    Yes, I think you can make a call to AIX Support for an explanation and perhaps advice on balancing the devices. But this may not be a performance problem, the higher level MPIO drivers may well balance out across the adapters.

    Note: nmon questions on the Performance Tools forum please.
    #AIX-Forum