This is what we use before we put our system into restricted state for a SavSys

File Layout
     A          R DPSYSRS                 TEXT('QINTER & QBATCH  ')
     A             RSFL1          6          TEXT('BLANK')
     A                                           COLHDG('BLANK')
     A            RSSBS         10         TEXT('SUBSYSTEM')
     A                                           COLHDG('SUBSYSTEM')
     A            RSFL2         27          TEXT('BLANK')
     A                                           COLHDG('BLANK')
     A            RSSBA          7          TEXT('JOBS ACTIVE')
     A                                           COLHDG('JOBS ACTIVE')
     A            RSFL3         82          TEXT('BLANK')
     A                                           COLHDG('BLANK') 

CLLE Program List
             PGM
             DCL        VAR(&CNTBA) TYPE(*CHAR) LEN(7) VALUE('*ZERO')
             DCL        VAR(&CNTIA) TYPE(*CHAR) LEN(7) VALUE('*ZERO')
             DCL        VAR(&JOBNME) TYPE(*CHAR) LEN(10)
             DCL        VAR(&USER) TYPE(*CHAR) LEN(10)
             DCL        VAR(&JOBNBR) TYPE(*CHAR) LEN(10)
             DCLF       FILE(DPPSYSRS)
             RTVJOBA    JOB(&JOBNME) USER(&USER) NBR(&JOBNBR)
             CLRPFM     FILE(DPPSYSRS)
             /**************************************************/
             OVRPRTF    FILE(QPDSPSBS) HOLD(*YES)
             WRKSBS     OUTPUT(*PRINT)
             CPYSPLF    FILE(QPDSPSBS) TOFILE(DPPSYSRS) +
                          JOB(&JOBNBR/&USER/&JOBNME) SPLNBR(*LAST)
             MONMSG     MSGID(CPF3303 CPF3311 CPF3429)
             DLTSPLF    FILE(QPDSPSBS) JOB(&JOBNBR/&USER/&JOBNME) +
                          SPLNBR(*LAST)
             MONMSG     MSGID(CPF0000)
 READ:   RCVF       RCDFMT(DPSYSRS)
             MONMSG     MSGID(CPF0864) EXEC(GOTO CMDLBL(END   ))
             IF         COND(&RSSBS *EQ 'QBATCH') THEN(DO)
             IF         COND(&RSSBA *NE '0') THEN(CHGVAR VAR(&CNTBA) +
                          VALUE(&RSSBA))
             GOTO READ
             ENDDO
             IF         COND(&RSSBS *EQ 'QINTER') THEN(DO)
             IF         COND(&RSSBA *NE '0') THEN(CHGVAR VAR(&CNTIA) +
                          VALUE(&RSSBA))
             GOTO READ
             ENDDO
             GOTO READ
END:
             SNDMSG     MSG('There are still ' |> &CNTIA |< ' users +
                          signed on and ' |> &CNTBA |< ' batch jobs +
                          running that have to be cancelled before +
                          terminating to a restricted state.') +
                          TOMSGQ(QSYSOPR)

             /* you could email or text out at this point */
             ENDPGM                                    

You can use the List Job API: QUSLJOB to do this. Running an API is generally faster way of doing things like this than running a command to produce a spool file, then reading the spool file.

Forget about the performance aspect. API's are upwards compatible with new releases. Format of a spooled file can change at any new release, possibly by PTF, meaning you'd have to review and possibly modify your program all the time.

But, just curious, are you not using save-while-active? The only thing I ever remember having problems with was a lock that prevented a *DTAQ from being saved. Never had troubles with open *FILE's, and, of course, there are no locks on active *PGM's.

(I eventually ended the QRCVDTAQ job right before the save, and restarted after giving a few minutes for the save-while-active checkpoint to be reached. Another option, in this case, would have been to omit the object, since there were no recovery issues.)

the reason I don't do save while active is a want a snapshot that is 100% "pure", not with some parts partially changed. However, thinking about it, maybe that is what I need to do and not worry about it.

I took at look at save while active on the savlib command and there are *NO, *SYNCLIB, *LIB and *SYSDFN and I read the text and I guess that *SYSDFN would be o.k. but if someone has a file open and in an update mode (which would be a record lock situation) is the SAVLIB going to sit? (if so then "save while active" doesn't help me.

To see how many jobs are active in a certain subsystem you could use the Retrieve Subsystem API (QWDRSBSD). See supplied code for an example.

Regards,

Carel Teijgeler

PGM PARM(&QSBS &RTNLIB &STATUS &ACTJOB)

INPUT: DCL VAR(&QSBS) TYPE(*CHAR) LEN(20)
DCL VAR(&RTNLIB) TYPE(*CHAR) LEN(10)
DCL VAR(&STATUS) TYPE(*CHAR) LEN(10)
DCL VAR(&ACTJOB) TYPE(*DEC) LEN(5 0)

JOB: DCL VAR(&RTNVAR) TYPE(*CHAR) LEN(360)
DCL VAR(&RTNLEN) TYPE(*CHAR) LEN(4)
DCL VAR(&DTAFMT) TYPE(*CHAR) LEN(8) +
VALUE('SBSI0100')
DCL VAR(&APIERR) TYPE(*CHAR) LEN(4) +
VALUE(X'00000000')

FOUTJE: DCL VAR(&KEYVAR) TYPE(*CHAR) LEN(4)

DCL VAR(&MSGDTA) TYPE(*CHAR) LEN(200)
DCL VAR(&MSGF) TYPE(*CHAR) LEN(10)
DCL VAR(&MSGFLIB) TYPE(*CHAR) LEN(10)
DCL VAR(&MSGID) TYPE(*CHAR) LEN(7)

MONMSG MSGID(CPF0000) EXEC(GOTO CMDLBL(ERROR))

CHGVAR VAR(%BIN(&RTNLEN)) VALUE(360)

CALL PGM(QWDRSBSD) PARM(&RTNVAR &RTNLEN &DTAFMT +
&QSBS &APIERR)

CHGVAR VAR(&RTNLIB) VALUE(%SST(&RTNVAR 19 10))
MONMSG MSGID(MCH3601)

CHGVAR VAR(&STATUS) VALUE(%SST(&RTNVAR 29 10))
MONMSG MSGID(MCH3601)

CHGVAR VAR(&ACTJOB) VALUE(%BIN(&RTNVAR 73 4))
MONMSG MSGID(MCH3601) 

RETURN

ERROR: RCVMSG MSGTYPE(*EXCP) RMV(*NO) KEYVAR(&KEYVAR) +
MSGDTA(&MSGDTA) MSGID(&MSGID) MSGF(&MSGF) +
MSGFLIB(&MSGFLIB)
SNDPGMMSG MSGID(&MSGID) MSGF(&MSGFLIB/&MSGF) +
MSGDTA(&MSGDTA) TOPGMQ(*PRV) +
MSGTYPE(*ESCAPE) KEYVAR(&KEYVAR)
MONMSG MSGID(CPF0000)

ENDPGM


*****************************************************************************************

CMD PROMPT('Retrieve subsystem information')

PARM KWD(SBS) TYPE(Q1) MIN(1) PROMPT('Name +
subsysteem')

Q1: QUAL TYPE(*NAME)
QUAL TYPE(*NAME) DFT(*LIBL) SPCVAL((*LIBL) +
(*CURLIB)) PROMPT('Library')

PARM KWD(RTNLIB) TYPE(*CHAR) LEN(10) RTNVAL(*YES) +
PROMPT('CL-var for RTNLIB (10)')

PARM KWD(STATUS) TYPE(*CHAR) LEN(10) RTNVAL(*YES) +
PROMPT('CL-var for STATUS (10)')

PARM KWD(ACTJOB) TYPE(*DEC) LEN(5 0) RTNVAL(*YES) +
PROMPT('CL-var for ACTJOB (5 0)')  

Record lock shouldn't hurt you, unless you're using commitment control, in which case there's a parameter related to how long to wait for commit boundaries. Unless you've got extremely long running commit cycles, you should still be able to reach a checkpoint in a reasonable amount of time (seconds or minutes).

You may want to review Systems management > Backup and recovery > Backing up your system. Basically all objects are checkpointed, which is the time where they're synchronized. There are several levels of synchronization, which you need to set based on the dependencies between your objects in different libraries. From the checkpoint time until the end of the save, for any storage pages that are changed, the system makes a copy of the storage page. The live page continues with the object. The copy of the page, which is the before version (what it looked like at the time of the checkpoint) is what gets saved. When the save is done, the "before" copies are destroyed, since they were only needed for the save.