Re: [SLE] Thanks. Re: Re: [SLE] How to get disk IO information?

From: Randall R Schulz (
Date: 09/07/05

  • Next message: Patrick Shanahan: "Re: [SLE] Could someone explain it for me?"
    Date: Tue, 6 Sep 2005 18:58:43 -0700


    On Tuesday 06 September 2005 13:10, raymond raymond wrote:
    > Hi, there are some people answered my question. Thanks lots.
    > But I still can't find the information what I need. So, here, I would
    > like to say some details about my problem.
    > I am using a test software (DOTs) to test databases performance.
    > It's is only supported on linux platforms.

    Would that be Database Opensource Test Suite

    > It need to accesses the
    > system level files (/proc/meminfo, /proc/stat etc.) and gets CPU,
    > disk and memory usage information by parsing the system level
    > files. I looked into the source code where it reads and parse the
    > system level file. I found it read /proc/stat file to get cpu, disk
    > and page information. It is supposed that in the /proc/stat file,
    > there are some lines start with "cpu", "disk" or "disk_io", "page",
    > and the program get the correspoding information from those lines.
    > But in Suse 9.3(the version I am using), I used "cat stat" to show
    > the content of the stat file, there are not disk_io and page_in_out
    > information. So far, I found the page_in_out information in the
    > vmstat file. But I still can not find the disk_io information.I
    > searched all of the file under /proc/ folder, but no resule.

    Which kernel are you running? For which kernel is the monitoring
    software written? Perhaps they're not the same and the particular
    information that software seeks is no longer in the place in which it
    was found in the kernel for which it was written.

    According the the DOTS page I found, the requirement is for kernel 2.4.4
    "or later." But I wonder if they've tested it on 2.6 kernels? I wonder
    if the pertinent introspection files have changed from 2.4.4 to the
    2.6.x kernels?

    This is why I recommended that you use applications to access this
    information. Their output tends to be more stable in the face of
    changes in low-level implementation details.

    If you have the source code for the monitoring software, you could
    modify it to accommodate the current information presentation in /proc
    and /sys.

    > The result of the test software I am tring to use is supposed to be
    > something like following:
    > ...
    > <Disk IO> Peak of this Interval: 430/s
    > Average of this Interval: 339/s
    > Peak of all: 430/s
    > Average of all: 339/s
    > <Page In> Peak of this Interval: 0/s
    > Average of this Interval: 0/s
    > Peak of all: 0/s
    > Average of all: 0/s
    > ...
    > Now, the disk_io is just 0, since the program can not get the
    > information. Someone could help me?

    It rather looks to me like it's reporting non-zero I/O numbers. Paging
    is zero, but that just means you're not paging, presumably.

    > ...
    > Thanks.
    > Raymond

    Randall Schulz

    Check the headers for your unsubscription address
    For additional commands send e-mail to
    Also check the archives at
    Please read the FAQs:

  • Next message: Patrick Shanahan: "Re: [SLE] Could someone explain it for me?"