HP Support Information Digests

===============================================================================
o IT Resource Center World Wide Web Service
   ---------------------------------------------------

   If you subscribed through the IT Resource Center and would
   like to be REMOVED from this mailing list, access the
   IT Resource Center on the World Wide Web at:

     http://www.itresourcecenter.hp.com/

   Login using your IT Resource Center User ID and Password.
   Then select Support Information Digests (located under
   Maintenance and Support). You may then unsubscribe from the
   appropriate digest.
===============================================================================


Digest Name: monthly HP-UX technical tips digest
    Created: Sat Jun 15 6:10:03 EDT 2002

Table of Contents:

Document ID Title
--------------- -----------
4000027449 What has changed in memory management between HP-UX 11.0 and 1
4000025970 SYS ADM: value of inod-sz is at maximum as reported by "
3200660306 HP-UX: How to recreate an /etc/fstab file
2200270845 SYS ADM: <b>swapinfo(1M)</b> shows only 2 GB swap on vg0

The documents are listed below.
-------------------------------------------------------------------------------


Document ID: 4000027449
Date Loaded: 20020517
      Title: What has changed in memory management between HP-UX 11.0 and 11.11 (11i)?

PROBLEM

It appears that memory management has changed between HP-UX 11.0 and
11.11. Global Memory metrics reported by Glance and MeasureWare vary
widely. For example, in HP-UX 11.0, GBL_MEM_USER includes "some kinds
of dynamically allocated kernel memory". (Description is taken from
the HELP page.

On HP-UX 11.11 (11i) and beyond, GBL_MEM_USER does NOT include some
kinds of dynamically allocated kernel memory. This is now reported
in the GBL_MEM_SYS* metrics.

On similarly configured systems the following differences appear with
MeasureWare C.03.35:

                             11.0 11i

      GBL_MEM_PHYS 1024 mb 1024 mb
      GBL_MEM_SYS 84.6 mb 310.6 mb
      GBL_MEM_USER 553.4 mb 191.7 mb

The value GBL_MEM_SYS on 11.11 is huge compared to 11.0, then the
reverse is true for GBL_MEM_USER.

What are the differences in memory management between HP-UX 11.0 and
11.11 (11i)?

CONFIGURATION

Operating System - HP-UX
Version - 11.11 (11i)

RESOLUTION

Generally, the kernel dynamically allocates and deallocates memory
while it is running. This is analogous to the user-space malloc(3)
routines, which dynamically allocate and free memory as needed in a
user process. The mechanism changed between HP-UX 11.00 and 11.11
to waste less memory in 11.11 and to aid in debugging.

To oversimplify slightly: the pre-11.11 algorithm would allocate all
kernel dynamic memory in sizes which were powers of 2 (32, 64, 128-
byte, etc), and all subsystems would share the same common allocation
pool. The 11.11 algorithm allocates from "arenas" in a variety of
sizes, and attempts are made to keep memory freed by one system within
that subsystem's arena, so as to avoid immediate re-use by another
subsystem.

As to the differences in Glance and MeasureWare reports, in 11.00, the
memory allocator did not implement a way to notify the KI (Kernel
Interface) based tools (midaemon, and, by extension, Glance) of memory
that had been allocated dynamically on behalf of the kernel. All
memory which could not be accounted for on 11.00 as either part of the
static kernel allocations (done at boot time), buffer cache, or free was
considered "USER" memory. So, the dynamic kernel allocations described
above would (incorrectly) be accounted-for as USER memory.

In HP-UX 11.11 (11i), the memory allocator implemented "hooks" to tell
KI tools how much memory has been dynamically allocated by the kernel,
and so this memory is now (correctly) accounted for as SYStem usage.
As noted above, this meant that a large amount of memory allocation got
reclassified from the USER to the SYS totals.

<!-- ALT KEYWORDS -->
-----End of Document ID: 4000027449------------------------------------------


Document ID: 4000025970
Date Loaded: 20020517
      Title: SYS ADM: value of inod-sz is at maximum as reported by &quot;sar -v&quot;

PROBLEM

In an HP-UX 11.0 environment, the value of inod-sz as displayed
by the "sar -v" command is confusing. The inode cache reaches
its maximum (or exceeds it if it can) several times a day.

Why does this symptom occur?

CONFIGURATION

Operating System - HP-UX
Version - 11.0

RESOLUTION

Generally, "inod-sz" is the number of active inodes in the system
including both HFS inodes and JFS inodes. When the number of
active inodes surpasses the kernel parameter "ninode", then
inod-sz will be bound with ninode's value.

Here are more details: Under the "inod-sz" field, sar gives two
numbers. According to the man page sar(1M), the first number is
the current size of the inode cache while the second one is the
maximum size of the inode cache.

The maximum size under the "inod-sz" field reflects the maximum
size of the HFS inode cache, as specified with the kernel parameter
"ninode". The current size, however, is evaluated as the sum of
BOTH HFS and JFS inodes that are active in the system. When the
sum of active HFS and JFS inodes surpasses the maximum size, i.e.
"ninode", the current size displayed will be bound with ninode's
value.

<!-- ALT KEYWORDS -->
-----End of Document ID: 4000025970------------------------------------------


Document ID: 3200660306
Date Loaded: 20020517
      Title: HP-UX: How to recreate an /etc/fstab file

PROBLEM

How can the /etc/fstab be re-created after it has been lost,
destroyed, or for whatever reason, does not exist?

CONFIGURATION

Operating System - HP-UX
Version - 10.X, 11.X

RESOLUTION

One method of re-creating an /etc/fstab files is as follows:

     1. # mount -p > /etc/fstab

     2. Verify that the permission, owner and group are:
         -rw-r--r-- root sys

         For example:

         # chmod 644 /etc/fstab

         # chown root:sys /etc/fstab

<!-- ALT KEYWORDS -->
-----End of Document ID: 3200660306------------------------------------------


Document ID: 2200270845
Date Loaded: 20020517
      Title: SYS ADM: &lt;b>swapinfo(1M)&lt;/b> shows only 2 GB swap on vg00 when 4 GB were d

PROBLEM

After defining a 4 GB swap partition on a vg00 disk, the "swapinfo -tam"
command only shows 2 GB. Upon the initial reboot, an error occurred
during swap allocation. The advice was to increase maxswapchunks by 1039,
and reconfigure system. After increasing this kernel parameter, the system
booted ok, but it still only showed 2 GB swap configured.

How can this problem be resolved?

CONFIGURATION

Operating System - HP-UX
Version - 11.0

RESOLUTION

Two factors contributed to this problem:

1. The maxswapchunk kernel limit was set too low and needed to be
    increased.

    The kernel tunable maxswapchunks is used to determine the amount of
    swap space that is configurable on the system. Use this formula to
    determine the maximum amount of configurable swap space:

     Max Swap Space = maxswapchunks * 1024 * swchunk (typically 2048)

     If maxswapchunks is 256 (default) then 512Mb of swap space can be
     configured.

          256 * 1024 * 2048 = 536870912

          536870912 / 1024 = 524288 Kb

          524288 / 1024 = 512 Mb

     Maximum Configurable Swap Space: 512 Mb

     Here is a second example of computing the maximum configurable swap
     space. In this example maxswapchunks is set to 2005.

          2005 * 1024 * 2048 = 4204789760 bytes

          4204789760 / 1024 = 4106240 Kbytes

          4106240 / 1024 = 4010 Mbytes

          4010 / 1024 = 3.916 Gbytes

          Maximum Configurable Swap Space: 3.9 Gb

     Use the sysdef(1M) command to find out what is configured on the
     system. For example:

          # sysdef
          ...
          maxswapchunks 512 - 1-16384 -
          maxtsiz 16384 - 0-655360 Pages -
          maxtsiz_64bit 262144 - 256-1048576 Pages -
          maxuprc 75 - 3- -
          ...
          swapmem_on 1 - - -
          swchunk 2048 - 2048-16384 kBytes -
          ...

2. Output from swapinfo(1M) indicated that the device still thought it
    had a filesystem on that disk even though it was configured as device
    swap in the fstab file.

    # swapinfo -tam

               Mb Mb Mb PCT START/ Mb
    TYPE AVAIL USED FREE USED LIMIT RESERVE PRI NAME

    dev 768 51 717 7% 0 - 1 /dev/vg00/lvol2
    dev 900 51 849 6% 0 - 1 /dev/vg00/newswap
    dev 2080 51 2029 2% 1966080 - 1 /dev/vg00/swapdev2
                                         ^
                                         |
                                         |--> normally zero

    From the swapinfo(1M) man page:

    START/LIMIT For device paging areas, START is the block address on the
                mass storage device of the start of the paging area. The
        ---> value is normally 0 for devices dedicated to paging, or
                the end of the file system for devices containing both a
                file system and paging space.

                For file system paging areas, LIMIT is the maximum number
                of 1-Kbyte blocks that will be used for paging, the same
                as the limit value given to swapon.

    Also, the following message occurred in /etc/rc.log:

         /usr/sbin/swapon: Device /dev/vg00/swapdev2 contains a file system.
         The system will page in the space after the end of the file system.
         Add "end" to the options field in /etc/fstab to turn off this warning.

    Perform the following steps to correct this problem:

          a. Since this is a device swap, it cannot be unmounted.
              Remove the entry from fstab, and reboot the system.

          b. Once the system is up, use dd(1) to copy a large file
              to that lvol, for example:

              # dd if=/stand/vmunix of=/dev/vg00/swapdev2

          c. Configure the swap, for example:

              # swapon -f /dev/vg00/swapdev2

          d. Check swapinfo as above. If everything looks ok then
              add the entry back into fstab.

<!-- ALT KEYWORDS -->
-----End of Document ID: 2200270845------------------------------------------

• Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]