install – ColinPaice

Some products and packages that run in Unix System Services provide .pax files you download and install. Often the instructions are not very clear.

The steps are usually:

Download the .pax file to your work station
Upload the file to z/OS ( usually Unix Services, but a sequential data set can be used.
Optionally create a ZFS file system, or find space on an existing file system
Unload the file

Download the .pax file to your work station

This is usually pretty simple – just make sure you are downloading the correct file. I spent a short while trying to get a .tar.gz file for Linux installed.

Upload the file to z/OS ( usually Unix Services, but a sequential data set can be used.

Upload the file in binary
I usually upload it into a zfs. This means you do not need to allocate space, and DCB information.

Optionally create a ZFS file system, (or find space on an existing file system)

This is where it starts to get harder – as there is less useful information. The first question is how much space do you need? Sometimes you get information in (mega) bytes, sometimes in number of 512 byte records, sometimes in number of 8KB blocks.

If you need to create a ZFS, you need to create a directory on the file system where you want the file system mounted, for example

mkdir /usr/lpp/java/J11.0.19.0_64

My pax file was in /tmp/ibm-semeru-certified-jdk_s390x_zos_11.0.19.0.pax.Z.

You can list the contents of the file using

pax -ppx -Evzf /tmp/ibm-semeru-certified-jdk_s390x_zos_11.0.19.0.pax.Z 1>aout 2>b

where aout has content like

drwxr-xr-x      1 JENKINS USERGRP     0 May 16 20:31 J11.0_64/lib/ 
-rwxr-xr-x apsl 1 JENKINS USERGRP 98304 May 16 20:13 J11.0_64/lib/default/libcuda4j29.so

Each line has extended attributes and the file size.

If the first character is “d” then this is a directory,
if it is a “-“, it is a file.
The sixth column is the size of the file.

You can use the following command to give the size of all of the files.

cat aout | awk ' substr($0,1,1) == "-" {print $0}' | awk '{sum+=$6;}END{print sum;}'

This command

passes the file aout into the awk command
awk… if the first character is a “-” =, it is a file, so print the record
awk…add up the 6th column and display the final result.

This gave me 5.24403e+08 (bytes) for the size of all of the files. You need to add perhaps 5-10% for overhead (for example directory entries). If you specify a secondary extend to the data set, it will try to expand if the ZFS fills up.

I created the ZFS

//IBMUZFS  JOB ,' ',COND=(4,LE) RESTART=MOUNT 
//DEFINE   EXEC   PGM=IDCAMS 
//SYSPRINT DD     SYSOUT=* 
//SYSIN    DD     * 
  DELETE               JVB800.V11.ZFS    CLUSTER 
  SET MAXCC=0 
  DEFINE                - 
    CLUSTER             - 
    (NAME(JVB800.V11.ZFS)- 
    VOLUMES(USER02)                   - 
    LINEAR              - 
    MEGABYTES(500 254)  - 
    SHAREOPTIONS(3 3)) 
/* 
//FORMATFS EXEC   PGM=IOEAGFMT,REGION=0M,COND=(0,NE,DEFINE), 
// PARM=('-aggregate JVB800.V11.ZFS    -compat') 
//SYSPRINT DD     SYSOUT=* 
//STDOUT   DD     SYSOUT=* 
//STDERR   DD     SYSOUT=* 
//*

and mounted it over the directory you created above

//MOUNT    EXEC PGM=IKJEFT1A,COND=((0,NE,DEFINE),(0,NE,FORMATFS))
//SYSTSPRT DD   SYSOUT=* 
//SYSTSIN  DD   * 
    MOUNT FILESYSTEM('JVB800.V11.ZFS') TYPE(ZFS) + 
    MOUNTPOINT('/usr/lpp/java/J11.0.19.0_64') + 
     MODE(RDWR) PARM('AGGRGROW') AUTOMOVE 
/*

You can use the MOUNT command in the BPXPRMxx member

    MOUNT FILESYSTEM('JVB800.V11.ZFS') TYPE(ZFS) 
    MOUNTPOINT('/usr/lpp/java/J11.0.19.0_64')
     MODE(RDWR) PARM('AGGRGROW') AUTOMOVE

Note: without the “+” signs.

Size after unpacking is similar to the calculated value above.

After the file was unpacked the command

df -P /usr/lpp/java/J11.0_64

reported

Filesystem     512-blocks   Used  Available Capacity Mounted on
JVB800.V11.ZFS 1178880   1045634     133246       89% /usr....

“Used blocks” 1045634 * 512 bytes gives 535364608 = 5.4e+08

The size calculation above is close to the final space used.

Unload the file

When I first unpacked one of the files I got messages about not being authorised to set the shared library attribute(l) in the directory – even though my userid was a super user. I had to define a security profile, and give my userid access to it.

RDEFINE FACILITY BPX.FILEATTR.SHARELIB UACC(NONE)
PERMIT BPX.FILEATTR.SHARELIB CLASS(FACILITY) ID(IBMUSER) ACCESS(READ)
SETROPTS RACLIST(FACILITY) REFRESH

Finally! In Unix System Services I issued the commands

cd /usr/lpp/java/J11.0.19.0_64
pax -ppx -rvzf /tmp/ibm-semeru-certified-jdk_s390x_zos_11.0.19.0.pax.Z

This unloaded the files into /usr/lpp/java/J11.0_64/J11.0_64

Discussion about where to put the file system.

If I had an existing directory /usr/lpp/java/J11.0_64 I could have mounted the new ZFS at this address, and issued

cd /usr/lpp/java
pax -ppx -rvzf /tmp/ibm-semeru-certified-jdk_s390x_zos_11.0.19.0.pax.Zf

and this would have unload the files into /usr/lpp/java/J11.0_64, overwriting what was already there. See Follow the instructions to install Java on z/OS and screw up production. Think carefully where you want to put your files.

Tag: install

Restoring pax files onto z/OS