AS/400 RUNBCKUP(*WEEKLY) Problem
We have seen a significant increase in the time it takes to run our weeklty backup which is a simply CL program calling RUNBCKUP(*WEEKLY). According to the log, it looks like it sits and does "nothing" for hours and hours. See the two log entries below:
Software/Hardware used:
CPC2197 Completion 00 05/24/08 00:23:25.289040 QLICHLBL QSYS 013C BK_WKL2004 SCCOSYS 0059
Message . . . . : Library DTM400V2 removed from library list.
Cause . . . . . : If the RMVLIBLE command was used, DTM400V2 was removed from the user library list. If the CHGSYSLIBL command was used, DTM400V2 was removed from the system portion of the library list.
CPC2198 Completion 00 05/24/08 00:23:25.319952 QLICHLLE QSYS 01A3 BK_WKL2004 SCCOSYS 0068
Message . . . . : Current library changed to *CRTDFT. Cause . . . . . : The current library in the library list was changed to *CRTDFT.
CPI6705 Information 00 05/24/08 10:07:45.513192 QTAERR QSYS 00EA QSRSVDEQ QSYS 0499
Message . . . . : 537 blocks processed for sequence 1, volume BACK01, on device TAP06.
Cause . . . . . : 537 blocks were processed for file label QFILEUPR, sequence 1 for volume BACK01 loaded on device TAP06. The maximum block size is 262144. The tape file was opened at 05/24/08 06:47:16 and closed at 05/24/08 10:07:45. The total time the file was open was 12029 seconds. If the device is a tape library the cartridge is *N and the resource is *N.
You can see it sits from 00:23 to 6:47 when it starts writing to tape. It is doing some processing. The reference to QFILEUPR indicates it is processing security data. My research so far leads me to think it has something to do with Private Authorties, but I was hoping someone out there had more insight or experience witht his. Software/Hardware used:
ASKED:
May 29, 2008 9:01 PM
UPDATED:
April 23, 2010 6:55 AM
Answer Wiki:
Hi,
That looks to me like 134Megs that it's writing to tape. That's an awfully long time to write that amount of data.
I would start by cleaning the tape drive and trying a brand new tape. I would also take a look at whether the job is waiting on locks while it's running. It may also be worth while taking a look at the qhst log and the problem log to see if there's anything else affecting your save.
Best regards,
Martin Gilbert.
=========================================================
The joblog is almost conclusive that this is indeed taking this much time with private authorities and probably has nothing to do with locks.
Review your authority scheme. Run PRTPRFINT *PCTFULL to see if any (non-IBM) profiles are collecting large numbers of authorities or if many profiles are all collecting authorities. You might review IBM profiles if ones that shouldn't be collecting authorities, e.g., QPGMR or QSYSOPR, are being involved in your site's objects. Some IBM profiles, e.g., QSYS will always have some large numbers.
If PRTPRFINT points to particular profiles, use DSPUSRPRF TYPE(*OBJAUT) to see what objects are involved.
If you have profiles that are out of line in terms of number of authorities, review how they become owners of or are privately authorized to objects. Break authorities into a set of profiles. Two profiles are better than one; three are better than two; but don't get carried away and move into complexity of management. Note that ownership can often be held separately from other authorities.
If authorities are spread of many profiles, then begin removing private authorities from individuals and assign to a set of group profiles. Then make individuals members of those groups. For applications, see if authority areas can be represented on a set of authority lists, and then place the group profiles on the lists after removing private authorities from them.
Getting a scheme started is the hardest part. There can be tens of thousands of authorities to assign. A few well thought out and straight-forward CL programs can go a long way towards handling most of the work.
Once a scheme is started, it should be easy to keep in place and to extend. Ask specific questions here as difficulties arise.
Tom
To see all answers submitted to the Answer Wiki: View Answer History.
I agree with Martin that you most likely are affected by object locks. I suspect the command runs with “save while active” options, and the defaults are 120 seconds to wait for each object is locked. I assume the runbkp command would do so as well. I’d also recommend that you display the tape to confirm that you are getting what you need backed up.
You might also consider creating your own cl with savsec, savlib, etc, so that is you need to end a subsystem you can do so that is timely for your applications.
Hi,
If this is a WEEKLY backup, should it not be in restricted condition to prevent locks and give the ablility to have a Full backup? When a backup is running and the objects are locked the system waits 120 seconds for each object just as AnchorS said.
TPinky