OS/390 R8 Upgrade

Hokey religions and ancient weapons are no
match for a good blaster at your side, kid.
--Hans Solo from Star Wars (ANH)

Here are my experiences upgrading to OS/390 R8. Please remember that problems I report here could most likely be from mistakes I made and not from product flaws.

Environment

IBM 9672-R46
From: OS/390 R6
To: OS/390 R8

Here's an adaptation of a quote for the R8 project:

My OS/390 R8 upgrade is like a game of chess, all of the mistakes are there waiting to be made.

Status

2000-02-18	Placed order for R8
2000-02-22	Started surveying ISVs for compatiblity with OS/390 R8

2000-03-06 - 2000-03-10	Attended the SHARE conference in Anahiem California. R8 User sessions: 2815 - OS/390 R8 User Experience given by Greg Thompson of US West 2873 - OS/390 Release 8 Experience given by Dan Phipps of Policy Management Systems Corporation Performed the pre-installation steps (I had a laptop with me at the conference). OS/390 R8 arrived while I was away (of course). Email trivia for the week: Home ISP: 40+ emails Work Internet: 100+ emails Work Intranet: 30+ emails
2000-03-11	Unpacked boxes, loaded the installation dialog and started reading the various installation manuals.
2000-03-14	Installed maintenance needed on my OS/390 R6 driving system as recommended in the Planning for Installation R8 manual.
2000-03-17	Started executing the installation dialogs.
2000-03-18	IPL'd "vanilla" system
2000-03-23	IPL'd with installation parms and products (like ACF2)

Our OS/390 R8 ServerPak order Trivia:

The PUT level is 9907
The RSU level is 9912
Came in 8 boxes
Physical Manuals: 69
Manuals for products we don't license: JES3 (4) and DFSORT (1)
OS/390 Installation Planning Kit marked "Start Here"
- Welcome to OS/390 Version 2 Release 8 sheet
- Hot Topics Newsetter R8
- Manuals:
  - Introduction and Release Guide R8
  - Information Roadmap R8
  - Planning for Installation R8
36 3480 compressed tapes (marked 210 MB):
- DLB01 - DBL31
- SERV01 - SERV04
- RIMS01
Tapes had create date of 02/27/2000
Other stuff:
- 4 library tapes
- 10 OS/390 Base source tapes
- OS/390 Licensed Product Library V2 CDROM
- 6 diskettes for LANRES and LAN Server
- PSF V310 came with 2 tapes and 5 manuals
- Java 118 came with 8 tapes
- GDDM came with 2 tapes

Pre-Installation tasks:

These are tasks that I know to do based on previous MVS and OS/390 installations:

HLQs defined to the security system:	OS390R8M - SSA OS390R8U - SSA SMPER8 - SMP for R8
DASD packs Initialized:	OSV28A - 3390-3 Catalog, IVP, etc OSV28B - 3390-3 SYSRES-1 OSV28C - 3390-3 SYSRES-2 OSV28D - 3390-3 HFS <- Needs SMS attribute OSV28E - 3390-3 DLIB 1 OSV28F - 3390-3 DLIB 2 These packs are NOT subject to archiving or incremental backups. We do weekly backups for Disaster Recovery. I do my own personal full volume backups at sync points when needed.
SMS ACS:	FILTLIST OS390R8O INCLUDE(OMVS.OS390R8.**) going to pack OSV28D

Analyzed my old OS/390 R6 SYSRES packs:

This is done so I place the R8 datasets on the right packs. For ease of install and backout, the master catalog will be okay with my either R6 or R8 SYSRES packs.

SYSRES-1 gets the traditional SYSRES files
SYSRES-2 gets new files and stuff I don't license

OSV26B - 3390-3 - SYSRES-1 - 89% used, HLQs:
```
API       APK       CBC       CEE       CPAC
EDF       EOY       EQAW      GIM       ICQ
IGY       ISF       ISP       PPFA      PSF
SCRIPT    SMPER6    SOMMVS    SYS1      TCPIP
```
Notes:
- SYS1 includes "legacy" like IPCS, RMF, VTAM, etc and JES2
- TCPIP datasets were the non-SMP datasets: ETC./, translate tables, etc.

OSV26C - 3390-3 - SYSRES-2 - 48% used - HLQs:

ANF       AOP       ASMA      ASMT      ASU
BFS       CSF       ECN       EHW       EPH
EUV       EWX       EZM       FFST      GDDM
GLD       ICA       IMW       IOA       IOE
JAVA      NETVIEW   PLI       PPMP      REXX
SYSMVIEW  SYS1      TCPIP     TME10GEM  VSAPL
VSF2

Notes:

SYS1 includes products we don't have like HRF* (RACF) ICE* (DFSORT), NFS*, SBDT*, SIES*, SIOA*, SIAT* (JES3), SORT* (DFSORT - not catalogued here in production)

Some shops have their OS/390 install datasets prefixed by "SYS1". I use to do this in the past but with OS/390, I went to using the IBM HLQs. The primary benefit is that all of the IBM manuals match our dataset names.

ServerPak Installation Dialog notes:

Installed maintenence for driving R6 system as recommended in the Planning for Installation R8 manual.
Reviewed PSP for OS390R8 ServerPac
I unloaded the LOADRIM job from the RIM tape. It has ",LABEL=(xx,SL)," on the end of a long line (longer because of our HLQ for the installation dialog). Us TMS customers have to do some editing to move the LABEL= to another line so we can add ",,,EXPDT=98000" to it. There's only 9 lines so this is no big deal...
As I've done in the past, I copied 'SYS9.SYS.OS390R8.SCPPCENU(CPPCSAMP)' to userid.CLIST(OS390R8). All that is needed is a "c custompac.Qualifier SYS9.SYS.OS390R8 all". When I need to get into the installation dialog, I use the command %OS390R8. Since I never clean up, you can see the start dates of my previous OS installs:
```
   Name                VV MM  Created     Changed
. MVS43                01.00 96/02/01 96/02/01 14:25
. OS390R2              01.00 97/02/12 97/02/12 17:53
. OS390R5              01.00 98/10/13 98/10/13 17:02 Not put into production
. OS390R6              01.00 99/03/04 99/03/04 10:45
. OS390R8              01.00 00/03/16 00/03/16 16:30
```
Trivia: Don't remember why but the MVS43 member does this: address TSO "alloc file(cidtabl) da('sysesa43.ipo1.cidtable') shr reuse"
When I install a ServerPak, I always do the "Full" option, to wit:
F - Full System Replacement installs a complete new IPL-able standalone system including all SMP/E-maintained libraries, SMP/E environment, operational data sets, and CustomPac sample data sets. The supplied operational data sets must be merged with or replaced by production operational data sets before the new system is used in production.
I like bringing up a "vanilla" version of the new operating system and running the IVPs there and also again on my cloned version. I also keep this small system available for use as a "rescue" system.
Task: Modify System Layout
- New HLQs (from R6)
  - ASM (Was ASMA and ASMT in R2 and R6)
  - CMX
  - DCZ
  - GSK
  - HCM
  - IMO
  - TIVOLI
- Action item: SYS1.PARMLIB.OS390R8.CSW(LNKLSTxx/LPALSTxx) will need ASMA./ and ASMT./ changed to ASM./ and the libaries will need to be cataloged to &SYSR2.
- Oops. Looks like the PSF ACIF product wasn't ordered... I just sent an email to my IBM rep to order it. PSF is a new version (and more $$$$) and in the confusion it looks like we missed getting this option.
- Good, I remembered to change the "SMPE" HLQ to "SMPER8" this time. I won't have to go in later and "butcher" this change in (like with my R6 installation). This simple HLQ change is done via primary command "CH DSN SMPE. SMPER8.". The reason I do this is so I can easily get to both releases of the operating system during this transition period:
  - SMPER6 alias - points to SMPE environment for OS/390 R6
  - SMPER8 alias - points to SMPE environment for OS/390 R8
Task: Alias and SSA - Assign SSA OS390R8U to CATALOG.OS390R8U and SSA OS390R8M to CATALOG.OS390R8M
Action item: Need to create an alias of SMPER8 on all systems to CATALOG.OS390R8U. I'm okay on the driving system (where I'm doing the install) but the other systems will need these catalogs connected and the aliases OS390R8M and OS390R8U created.
Sometimes I feel like Charlie Brown expecting Lucy to hold the football so I can kick it. I miss this everytime I've installed OS/390: the user.linklib dataset SHOULD exist. The COPYICK step of job RESTORE says it is optional so no big deal.
The R8 SMPPTS is allocated at PRI=60480 SEC=6000 in blocks (6160). Job RECHOLD receives service tape PTFs. We have our Sterling (now CA) product VAM (SAMS:ALLOCATE) customized to display a message in the job log when a non-VSAM batch job has a dataset going to another extent. I was sweating it as I saw it in extent 10 as it called for the last tape but it ended up in 13 extents when the job ended. Action Item: Enlarge SMPER8.SMPPTS.
Job SYSALLOC didn't complete successfully because volume OSV28A didn't have enough space for the PAGE datasets. Moved SMPER8.SMPPTS to volume OSV28F and made it slightly larger. Found that job SYSALLOC doesn't restart as expected. I ended somehow with the PAGE.CPAC.* datasets that couldn't be deleted so I created my own job to create PAGE.CPAC2.* datasets (with a 1 cylinder PLPA.
Job DMPBLD appeared to do what I wanted but the delete of the modules in step DELAUTH failed (maybe because the COPYICK step of the RESTORE job didn't run???).
I elected not to run job CPYVTLCL. Instead of putting a lot of crap in CPAC.VTAMLST, I created a new proc called "NET" in CPAC.PROCLIB that concatenates our VTAMLST and VTAMLIB in the proc copied from CPAC.PROCLIB(VTAM). The VTAMLIB datasets were added as APF entries in CPAC.PARMLIB(PROG00) The installation VTAMLST and VTAMLIB datasets were copied to volume OSV28A so other packs aren't needed in case I need this as a "rescue" system.
At this time I created the member CPAC.PROCLIB(SHOWMVS) and added its steplib APF entry in CPAC.PARMLIB(PROG00). Part of my personal IVP is to run SHOWMVS (the great MVS trivia program - from file 183 of the http://www.cbttape.org). on the vanilla R8 system for comparison:
```
//SHOWMVS  PROC
//JUSTDOIT EXEC PGM=SHOWMVS,TIME=1440,REGION=6M
//STEPLIB  DD DISP=SHR,DSN=userid.DAVE.LOAD,
//            UNIT=3390,VOL=SER=TSO006
//SHOWMVS  DD SPACE=(CYL,(2,2),RLSE),UNIT=3390,VOL=SER=OSV28F,
//         DSNAME=SYS9.SYS.OS390R8.SHOWMVS,
//         DISP=(NEW,KEEP)
```

I also created member CPAC.PROCLIB(XTRDAVE) for writing the SYSLOG to DASD:

//XWTR    PROC
//IEFPROC EXEC PGM=IASXWR00,REGION=64K,
//             PARM='PL',TIME=1440
//IEFRDER  DD SPACE=(CYL,(1,1),RLSE),UNIT=3390,VOL=SER=OSV28F,
//         DSNAME=SYS9.SYS.OS390R8.SYSLOG,
//         DISP=(NEW,KEEP),
//         DCB=(BLKSIZE=13030,LRECL=137,BUFL=13030,BUFNO=2,RECFM=VBM)

IPL'd the "vanilla" OS/390 R8 system.
- IPL looks pretty normal except the wrong time offset. We haven't converted to GMT yet and CLOCK00 has a 6 hour offset. I ended up using a different IODF name other than SYS1.IODF00 and created a new LOADxx member in SYS1.IPLPARM to use it. I noticed that CPAC.PARMLIB and CPAC.PROCLIB were put on the master catalog volume instead of the RES volume as in the past. This is a better home for a user modified datasets (in the grand scheme of things).
- Message ILR005E PLPA PAGE DATA SET FULL, OVERFLOWING TO COMMON DATA SET appears as expected because my PLPA is at one cylinder. Funny, IBM recommends doing this for performance and then spits out a this error message twice in the IPL process.
- Even though we don't license RACF, it started automagically and then gripes because it isn't licensed. No harm done but I see message *BPXP006E OMVS IS WAITING FOR SECURITY PRODUCT INITIALIZATION Glad I didn't really need it (like to run TCP/IP)! Earth to OMVS: don't hold your breath.
- For some reason, IBM includes a printer at UCB=000F in the JES2 init statements. Since we don't have a printer at that UCB address, there's an hickup in the startup. I bypass the init statement with '/* */' and continue the startup.
- SYSLOG from this IPL
- SHOWMVS output from this IPL
- Now I'm ready to modify my CLONERES job to copy the vanilla SYSRES volumes to our standard IPL packs (with some modification). This job does this:
  - FDR Volume copy of SYSRES-1 volume OSV28B to RESxy1
  - A few misc non-SMP/E modifications are done to RESxy1
  - FDR Volume copy of SYSRES-2 volume OSV28C to RESxy2
  - A few misc non-SMP/E modifications are done to RESxy2
  - Selective FDR dataset copy from volume OSV28D:
    - OMVS.OS390R8.ROOT to OMVS.&sysid..&sysr3..ROOT on RESxy3
    - OMVS.OS390R8.ETC to OMVS.&sysid..&sysr3..ETC on RESxy3
  Legend: where "RESxyz" is:
  - x = LPAR number of 1 (production) 2 (Y2K/appl test) 5 (Sysprog test)
  - y = A or B set of RES volumes for this LPAR (that I flip flop between)
  - z = SYSRES 1 2 or 3
I changed job ESTABFS to use OMVS.OS390R8.* (instead of OMVS). Job RESTFS was also changed. These two jobs were executed many many times until I got the ACF2 security rules right for the BPX.FILEATTR.PROGCTL, BPX.FILEATTR.APF and BPX.SUPERUSER SAF rules coded correctly. Fun stuff.
Job RESTFS expects that CPAC.HFSFILE be cataloged. It was - to the one for R6. Luckily I researched this before submitting the job and recataloged it to volume OSV28A.

I found that 5 of the 37 NETVIEW datasets were missing the V1R2M0 dataset node. I added this jobstep to CLONERES, my SYSRES clone job, to allow my R8 system to work fine with existing APF entries, catalog and proc dataset names:

//FIXNETV EXEC PGM=IEHPROGM
//RES5B2   DD DISP=SHR,VOL=SER=RES5B2,UNIT=3390
//VDSBYPAS DD DUMMY
//SYSPRINT DD SYSOUT=*
//SYSIN    DD *
 RENAME DSNAME=NETVIEW.CNMLINK,                                        C
               NEWNAME=NETVIEW.V1R2M0.CNMLINK,                         C
               VOL=3390=RES5B2
 RENAME DSNAME=NETVIEW.NVULIB,                                         C
               NEWNAME=NETVIEW.V1R2M0.NVULIB,                          C
               VOL=3390=RES5B2
 RENAME DSNAME=NETVIEW.SCNMCRUN,                                       C
               NEWNAME=NETVIEW.V1R2M0.SCNMCRUN,                        C
               VOL=3390=RES5B2
 RENAME DSNAME=NETVIEW.SEKGMOD1,                                       C
               NEWNAME=NETVIEW.V1R2M0.SEKGMOD1,                        C
               VOL=3390=RES5B2
 RENAME DSNAME=NETVIEW.SEKGMOD2,                                       C
               NEWNAME=NETVIEW.V1R2M0.SEKGMOD2,                        C
               VOL=3390=RES5B2

Okay, I've found that the IBM-supplied SMF exits have moved from CPAC.SAMPLIB to SYS1.SAMPLIB and now have a prefix of IEE but where is SMFDUMP? For the time being, I'll use the one I got with R6.
I've IPL'd OS/390 R8 with our stuff: ACF2, CA-1, ABENDAID, SAMS:Allocate, ETFA, etc. I didn't really see anything out of the ordinary so far that wasn't caused by me shooting myself in the foot, like forgeting that my SYSRES-3 pack was freshly initialized so I DID need to copy the /etc file over this time. Dooh!

I like the new ISPF member lists with the full 4 digit year. It's both new and similar at the same time. That's probably an ISRCONFG option but I'll take the new format if given the choice. I know that ISRCONFG is done differently in ISPF v2.8 and its customization will come later.

... This project is on hold. The company that is buying us is installing OS/390 R8 and I plan to use their SYSRESs ...

Production!

Our merger is complete and we put this LPAR on the R8 SYSRES set from the new company. With a little manipulation, we were able to use their SYSRES volumes:

Catalog all datasets on the two SYSRES volumes to be prefixed with SYS1 where needed. For example:
- Old: CEE.SCEERUN
- New: SYS1.CEE.SCEERUN
Ensure our LE options were compatible -- CICS options were changed.
Override a DADSM exit here and SDSF user exit there.

All in all, R6 to R8 was pretty painless.