Tips and Tricks site for advanced HP-UX Engineers

01 Oct 07 Welcome to

This site was created for two reasons. It was meant to document procedures, scripts and processes that I felt were important to myself and the greater HP-UX community.

This project is owned by ISN Corporation, a Subchapter S Corporation based in Chicago, Illinois, USA.

If you want to contribute to the site, let me know. hpuxadmin in gtalk hpuxconsulting in yahoo messenger


Steven “Shmuel” Protter: Rosh Tzurim, Israel

Tags: , , ,

01 Dec 22 HP-UX system runs out of tty devices. To many sessions

  1. logins will hang, looking for a tty device
  2. kctune | grep npty
  3. kctune | grep nstrpty
  4. <removed>
  5. “PTY allocation request failed on channel 0” should be on the hang screen
  6. kctune npty=240 ; kctune nstrpty=240 

    A reboot will be required to implement the change.

19 Jul 22 Disk issues on HP-UX

So you are having disk problems:

Scenario 1:

You need to build a logical volume on a disk that is in the kernel. You have other issues that will not let you completely replace /etc/lvmtab

mv /etc/lvmtab /etc/

vgscan -av

You need that disk number 967 to change.

ioscan -NfnCdisk shows the offending device files




What to do?

Force the system to hand out new device files.

rmsf /dev/rdisk/disk967

rmsf /dev/disk/disk967

rmsf -H 64000/0xfa00/0x102

scsimgr -h lun_reset -D /dev/rdisk/disk967

insf -e

Possible alternative:

vgreduce -f /dev/vgname

Trick of the day if you are seing flapping on the SAN. Such as a scenario where deices are giong on and ofline.

ioscan -fN | wc -l ## Gets you a count.

drop the wc -l and you get a pretty detailed look at the HBA/SAN/disk infrastructure.

ioscan -fN

16 Dec 21 Linux vgs port for HPUX (updated to work with vg version 2.2)

A while back, I came across a port of the very nice linux vgs command for HP-UX. A h/t to Juan Manuel Rey ( for that original version.

But, upon using it recently, I noted it was not working now that we have some volume groups with vg version 2.2. Here is a modified version of the script that works for version 2.2 volume groups!

# - script to emulate the Linux LVM command vgs in HP-UX 11iv3
# (C) 2010 - Juan Manuel Rey (
# This program is free software; you can redistribute it and/or
# modify it under the terms of the GNU General Public License
# as published by the Free Software Foundation; either version 2
# of the License, or (at your option) any later version.
# This program is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# GNU General Public License for more details.
# You should have received a copy of the GNU General Public License
# along with this program; if not, write to the Free Software
# Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA  02111-1307, USA.

version="v0.2 2021/12/15"

usage() {
  echo "VGS for HP-UX ${version}"
  echo "Usage: vgs [-v vg_name]"
  exit 1

if [ ! "$(uname -r)" = "B.11.31" ]; then
   echo "VGS for HP-UX only works on HP-UX 11iv3"
   exit 1

if [ "$1" ]; then
"vgs" 87 lines, 3053 characters

if [ "$1" ]; then
   case "$1" in
   -v) shift; [  "$1" = "" ] && usage || vg_name=${1};;
   *)  usage;;

vg_display="vgdisplay -F"
[ ! "${vg_name}" = "" ] && vg_display="vgdisplay -F ${vg_name}"

printf "%-10s %-5s %-5s %-20s %-8s %-6s %-5s\n" VG PVs LVs Status Version VGSize Free

# First, we must determine which VG Version we have

eval ${vg_display} | \
  while IFS=":" read vgdisplay; do
     version=$(echo ${vgdisplay} | awk -F: '{for (i=1;i<=NF;i++) print $i}' | grep vg_version | cut -d= -f2)
     case "$version" in
       2.2) total_pe_index=15
         *) total_pe_index=13

     echo ${vgdisplay} | cut -d ":" -f 2 | cut -d "=" -f 2 | read status
     if [ "${status}" = "deactivated" ]; then
        echo ${vgdisplay} | cut -d ":" -f 3  | cut -d "=" -f 2 | read status
        echo ${vgdisplay} | cut -d ":" -f 3  | cut -d "=" -f 2 | read status
        echo ${vgdisplay} | cut -d ":" -f ${total_pe_index} | cut -d "=" -f 2 | read total_pe
        echo ${vgdisplay} | cut -d ":" -f ${pe_size_index} | cut -d "=" -f 2 | read pe_size
        echo ${vgdisplay} | cut -d ":" -f ${free_pe_index} | cut -d "=" -f 2 | read free_pe
        vg_size="`/usr/bin/expr $total_pe \* $pe_size / 1024`G"
        vg_free="`/usr/bin/expr $free_pe \* $pe_size / 1024`G"

     echo ${vgdisplay} | cut -d ":" -f 1 | cut -d "=" -f 2 | cut -d "/" -f 3 | read vg_name
     echo ${vgdisplay} | cut -d ":" -f ${cur_pv_index} | cut -d "=" -f 2 | read pvs
     echo ${vgdisplay} | cut -d ":" -f 5 | cut -d "=" -f 2 | read lvs
     #echo ${vgdisplay} | cut -d ":" -f 19 | cut -d "=" -f 2 | read version
     printf "%-10s %-5s %-5s %-20s %-8s %-6s %-5s\n" "${vg_name}" "${pvs}" "${lvs}" "${status}" "${version}" "${vg_size}" "${vg_free}"


16 Dec 21 “Subtracting” one file from another

I recently had the occasion to refactor a script (not mine) in which there was some convoluted logic to ensure that the contents of file A (an ascii file) were fully contained within file B (another ascii file). Lots of looping and grepping were the order of the day. I imagined that there had to be better way and first went down the path of using grep with a file as the source of the things I was looking for. However, while that would get me partially there, it would not tell me if File A was wholly contained with file B. I am sure I could made it work (somehow) using this idea, but again, I thought that there has to be a better way.

With the help of Google (because of course), I stumbled across an HP-UX command (not unique to HP-UX of course) that in my 25 years of scripting on the HP-UX platform, I had never encountered nor used before: the comm command. This command lets one implement set logic on ascii files.

Let’s have a look . . .

From the man page for comm:

      comm - select or reject lines common to two sorted files

      comm [-[123]] file1 file2

      comm reads file1 and file2, which should be ordered in increasing
      collating sequence (see sort(1) and Environment Variables below), and
      produces a three-column output:

       Column 1:   Lines that appear only in file1,
       Column 2:   Lines that appear only in file2,
       Column 3:   Lines that appear in both files.

  If - is used for file1 or file2, the standard input is used.

  Options 1, 2, or 3 suppress printing of the corresponding column.
  Thus comm -12 prints only the lines common to the two files; comm -23
  prints only lines in the first file but not in the second; comm -123
  does nothing useful.

So, “comm -12” performs the intersection operation on the files. But “comm -23” should do the trick for what I was after: Subtracting file B from File A. In my use case File B should be a proper subset of File A. I can test for that via:

checkCount=$(comm -23 $FILE_B_SORTED $FILE_A_SORTED | wc -l)

If the resulting count is “0”, I know that File B is wholly contained within File B. If the count is not “0”, then I can consume what is in File B but not in File A and take appropriate action.

Note that the two ASCII files need to be in sorted order – easy enough. Here is an example of all of this put together to accomplish of “subtracting File B from File A”:


sort -u $FILE_B > $FILE_B_SORTED
sort -u $FILE_A > $FILE_A_SORTED

# 'comm -23 file1 file2' says show me all the lines that appear in
# file2 but not in file1.  Thus, here, we are ensuring that there is
# nothing in the file2 that is NOT in file1.
# We count how many such lines there are - we expect there to be zero.

checkCount=$(comm -23 $FILE_B_SORTED $FILE_A_SORTED | wc -l)
if (( checkCount == 0 )); then
   echo "All entries in the File B are in File A  (goodness)"
   # Do something with the entries in File B that are not in File A
   comm -23 $FILE_B_SORTED $FILE_A_SORTED > $checkList
   exec  4< $checkList
   while read entry <&4; do
     # Do whatever is that needs to be done

All good stuff 🙂

Tags: , ,

17 Jun 20 HP-UX Serviceguard Missing node

If you have a hardware fault or other calamity in a HP-UX serviceguard cluster you lose the ability to make incremental changes to the cluster until that node comes back.

If you need to make a change to a cluster in this state and you don’t want to bring down the cluster, you have to do all your changes with one gigantic command line.

Lets say you have a 4 node cluster named cnode1,cnode2,cnode3, and cnode4.

cnode4 suffers a hardware fault and you packages fail over to cnode1-3. But your usage has grown and you have a package that is beating the hardware down and you want to move it from cnode2 to cnode1.

Well you can’t do it incrementally. You have to do it all at once. I recently ran into a situation where I had to modify 37 cluster environment files and the cluster configuration to remove a node cnode4.

That requires you to correctly type a command line that could easily be in excess of 4000 characters. Anybody who knows my typing skills knows this is beyond my abilities on my best day.

So I wrote a little assistant program.

It consists of three files two of which are scripts.

pkg-mod-list (A list of all the package configuration files, full path that need to be modified. It is your choice how to handle the editing. We used ansible last night when we did it in a DR cluster.

Contents …



Then we have helper scripts which put the command line together.

myclusterV6_prod.conf is the main cluster configuration file with the references to node cnode4 commented out.

cat missing-node-checkconf
MAIN=”cmcheckconf -C /etc/cmcluster/configs/myclusterV6_Prod.conf”
cat pkg-mod-list | while read -r pfile
PCMD=”${PCMD} -P ${pfile}”
### echo “$PCMD”
echo $MYCMD

exec ${MYCMD}

MAIN=”cmapplyconf -C /etc/cmcluster/configs/myclusterV6_Prod.conf”
cat pkg-mod-list | while read -r pfile
PCMD=”${PCMD} -P ${pfile}”
### echo “$PCMD”
echo $MYCMD

exec ${MYCMD}

13 Nov 19 Using File Descriptors other than stdin/stdout/stderr in Shell Scriptings

For a longish ime, I had to jump through major hoops to script around the issue of my standard input getting clobbered when inside a loop that was iterating over something that coming from stdin. I don’t have an exact example that recreates the issue but something like this would generate lots of headaches:

cat <some file> | awk '{<some fancy awk-type hingys>}' | \
  while read entry; do
    bla; bla; bla
    bla; bla; bla

Sorry I cannot provide an actual snippet of code to recreate the issue but what I would find in these situations is that my loop would end after one iteration (when I knew there should have been a lot more) and I coudn’t figure out why which made me haz a sad.

Well, I found a way to steer clear of all that by assigning a <some file> to a file descriptor different from stdin. To wit:

exec 3< /path/to/file
while read entry <&3; do
   bla; bla; bla
   bla; bla; bla

There are no doubt other ways to solve this conundrum but this is the way I have avoided it for quite some time now. Of course, make sure you do not associate a file with file descriptors 0,1,2 (unless you are quite sure that is what you want to do!).

13 Nov 19 Convert LVM volume group from 1.0 to 2.2

# First – show proof that we are currently using VG Version 1.0

# vgdisplay /dev/vgscott01
--- Volume groups ---
VG Name                     /dev/vgscott01
VG Write Access             read/write
VG Status                   available
Max LV                      255
Cur LV                      3
Open LV                     3
Max PV                      32
Cur PV                      16
Act PV                      16
Max PE per PV               40960
VGDA                        32
PE Size (Mbytes)            32
Total PE                    131204
Alloc PE                    65849
Free PE                     65355
Total PVG                   0
Total Spare PVs             0
Total Spare PVs in use      0
VG Version                  1.0   <----- Proof :-)
VG Max Size                 40t
VG Max Extents              1310720


# Step #1.  From experience, we *know* that each current disk in the VG will
# need to have at least one free extent at the end to accomodate the transition
# to VG Version 2.x.
# But, currently, all extents from all current disks are allocated.  
# We can move off 'N' number of extents from the end of the current 
# disk to a new disk using the "-e" option of the pvmove command.

# (note: we use dsf's for all commands)


# First we need to unmount all filesystems

umount /bb/9/dev/mylab
umount /bb/9/dev/venturelab
umount /bb/9/dev


# Ok, we need to free up at least one extent from the end of
# the original, fully-qualified disks.
# We use the "-e" option of the pvmove command to make this happen.

# pvmove -e 2 /dev/disk/disk247 /dev/disk/disk265
# pvmove -e 2 /dev/disk/disk248 /dev/disk/disk265
# pvmove -e 2 /dev/disk/disk251 /dev/disk/disk265
# pvmove -e 2 /dev/disk/disk249 /dev/disk/disk265
# pvmove -e 2 /dev/disk/disk260 /dev/disk/disk265
# pvmove -e 2 /dev/disk/disk259 /dev/disk/disk265
# pvmove -e 2 /dev/disk/disk258 /dev/disk/disk265
# pvmove -e 2 /dev/disk/disk261 /dev/disk/disk265
# pvmove -e 2 /dev/disk/disk262 /dev/disk/disk265
# pvmove -e 2 /dev/disk/disk263 /dev/disk/disk265
# pvmove -e 2 /dev/disk/disk264 /dev/disk/disk265

# At this point, a vgdisplay should show at least 2 extents for each physical volume. 
# Check our work

# vgdisplay -v /dev/vgscott01


# vgversion -r -V 2.1 /dev/vgscott01

# We are good to go!

# We need to deactivate the Volume Group before we can upgrade

# vgchange -a n /dev/vgscott01

# We are good to go!

# vgversion -V 2.1 /dev/vgscott01


# Check our work

# vgchange -a y /dev/vgscott01

# vgdisplay -v /dev/vgscott01


# Now, we can extend our logical volume

# lvdisplay /dev/vgscott01/lvol1

# lvextend -l 98864 /dev/vgscott01/lvol1
# fsadm -F vxfs -b 3163648m /bb/9/dev


# lvdisplay /dev/vgscott01/lvol1


# vgcfgbackup /dev/vgscott01


# mount /bb/9/dev
# mount /bb/9/dev/mylab
# mount /bb/9/dev/venturelab

# bdf /bb/9/dev

10 Nov 19 How to be a yes man

Learning something new is great. Joe Geiger taught me something cool that I should have learned years ago.

Serviceguard users ever wanted to script a cluster change such as a node add.

cmapplyconf -v -P <package file>

Ends with a y/n prompt do you want to apply? Normally that requires input. Not with the yes command:

cmcheckconf -v -P <package file>


# Check return code if not zero stop

if [ ${rc} -ne 0 ]


echo “Checkconf error ${rc}”

exit ${rc}


yes | cmapplyconf -v -P <package file>

# Check return code here as well

Tags: , , , ,

03 Sep 19 Add new dump device

add device called lvdump1


:/home/root # lvchange -r n /dev/vg00/lvdump1

Set dump lv to contiguous. SEP

lvcreate -L <size in MB> -n lvdump1 /dev/vg00

# lvchange -C y /dev/vg00/lvdump1
Logical volume “/dev/vg00/lvdump1” has been successfully changed.
Volume Group configuration for /dev/vg00 has been saved in /etc/lvmconf/vg00.conf

# crashconf -a /dev/vg00/lvdump1

:/home/root # crashconf -v

15 Aug 19 Turning of hardware test at boot on BL class Itanium blades

The Before picture:

BOOTTEST Settings Default Variable OS is not speedy boot aware. Selftest Setting

——— ————–

early_cpu Run this test

late_cpu Run this test

platform Run this test

chipset Run this test

io_hw Run this test

mem_init Run this test

mem_test Run this test

BOOTTEST Settings Default Variable OS is not speedy boot aware. Selftest Setting

Shell> boottest

Shell> boottest mem_test off

Shell> boottest mem_init off

Shell> boottest io_hw off

Shell> boottest chipset off

Shell> boottest platform off

Shell> boottest late_cpu off

Shell> boottest early_cpu off

BOOTTEST Settings Default Variable OS is not speedy boot aware. Selftest Setting

——— ————–

early_cpu Skip this test

late_cpu Skip this test

platform Skip this test

chipset Skip this test

io_hw Skip this test

mem_init     Skip this test

mem_test Skip this test

Skip this test Shell>

This cuts boot time over 90%. NOTE: YHou may need to turn this stuff back on to troubleshoot hardware issues.

WhatsApp chat