I want to share an experience that I had over the last two weeks-we lost a disk drive and at the same time found out that our backups were bad.
Four years ago I wrote the programs to automate backup, but we had never tested them. My programs ran a SAVSYS to tape (which took one full tape and a portion of a second tape), immediately followed by a SAVLIB *NONSYS on the same second tape. I found that, although both commands were running fine, the SAVLIB *NONSYS was overwriting the end of the SAVSYS.
Keep your SAVSYS separate from your SAVLIB *NONSYS! It will prevent headaches. It took over 2 1/2 days to recover from this disaster and countless hours with IBM Levels 1 and 2 over the weekend.
Print out your user profiles and your device descriptions! Having them on tape is no good if you can't read the tape, since the system will only read the tape if it has an operating system. Your work is totally in the blind.
Check your tapes regularly! It is very important to make sure your backup plan is really working out. Exercise it if you have the luxury of time and resources.
Note From Tim Johnston:
Keep a SAVSYS, SAVLIB *IBM and a SAVLIB *NONSYS in a fireproof safe (small one just for tapes and diskettes).
Daily, do a RTVCFGSRC to a CL source file. That way, even if things go totally awry, you'll have the descriptions for all your devices, lines, etc.
Daily, do a SAVCHGOBJ on all of your production libraries. In that command, specify ENDOPT (*LEAVE).
Then do a SAVSECDTA to the tape, again with ENDOPT(*LEAVE).
Lastly, you can use SAVSPLF (MC, August 1991, by yours truly) against all existing spool files. Following suit, get rid of all joblogs.
My SE and a business partner with whom I work assure me that this is a sound backup plan.
Note From Dennis Detwiler:
I found a weak link when libraries were created between SAVCHGOBJ runs. The SAVCHGOBJ command doesn't do anything with new libraries until the library is saved for the first time with SAVLIB. I suggest you do a SAVLIB on any newly created libraries.