Tape backup speed - Erratic

[Former Member]

Hi,

We have a 570 backing up a 3TB 4.7 Unicode database (save while active) to a 3592 tape drive. Some weeks the backup takes 12 hours. Other weeks the backup runs 24 hours or anywhere in between. We have not seen any appreciable difference in the system load at the time the backups are running. Has anyone else seen anything like this? It's really puzzling.

Thanks,

Craig

[Former Member]

Could you show the exact command with all its parameters that you used?

Are you using internal or external disk subsystem? If external, do you dedicate entire ranks to i570? If not, you need to check if the shared ranks are being accessed with high IOs from other machine/LPAR while you do the save on i570.

Do you notice that the same set of tape cartridges are used when the save time exceeds 12 hours? This can indicate deteriorating tape media quality. To display tape error statictics:

- STRSST and provide correct service tool user/password

- Select 1 Start a service tool

- Select 1 Product activity log

- Select 4 Work with removable media lifetime statistics

- Specify proper "Removable media" (most likely = 2)

- Leave "Sort By" as 4 (All Error Rate)

You should then see a list of tape error rate with the Volume ID displayed.

Next time when the job running the save command runs more than 12 hours, please also try using WRKJOB command to look at its job log to see if there is any suspicious messages that can provide some more clues.

Satid S.

IBM Thailand

[Former Member]

Hi Craig --

We had something similar happening where all of a sudden we went from 4 hrs to 6 hrs on a TS3310 LTO tape library.

After spending numerous hrs / intermittent days looking at this without success; in the mean time a disk drive failed and the next day after replacing the drive; our backups were back down to 4 hrs.

Not that this helps you, but I think there's something odd with degradation with backups to tape when a drive is on it's way to failure. (I know this sounds far fetched!) You would think the performance hit would be across the board (effecting SAP too) but it wasn't.

Perhaps IPL the server? Our drive finally reported in as failed during an ipl

Good Luck

[Former Member]

Hi Craig,

Have a look in your system log using DSPLOG or DSPMSGQ QSYSOPR.

You might be seeing some messages where it is trying to do a save while active and there are either pending record changes or it is waiting for commitment control.

We tend to have problems with save while active and therefore do off-line backups. If you have not the luxury of getting any downtime maybe think about mirroring.

Failing that I would ask IBM for assistance.

Gary

[Former Member]

Hi Craig,

If you are doing a online save, then it may be happening that on some of the weekends, it is able to get a Object lock sooner than the other weeks, quite a remote possibility...but nevertheless worth taking a look at...Are there a lot of jobs running on the weekends..at that time??

Also for the parameter "Pending record changes " are you using a value or are you using *NOCMTBDY?? Maybe that value is too high..

Apart from that, I do not have much clue...probably ask IBM, if u are using BRMS or your hardware vendor.

Thanks

Abhi