Hello All,

We have a problem in one of the partitions for IBM i. Daily FULL Backup is taking more time each day. On Mondays backup is taking the lowest time, while on Saturdays it is taking the longest time. IPL is taking place on Sundays either after weekly or monthly backup. Backup is being performed on VTL tapes.

Backup is using the same Control Group and a FULL backup is done each time. Also backup is done in a non restricted state but all application subsystems are powered down before the backup.

I have checked data being backup-ed up, it is approximately the same size each day. 

Any idea as where to check in the system to know the cause of this problem ?

Thanking you in advance,

Nahla Alloush
Advisory IT Specialist

------------------------------
Nahla Alloush
------------------------------

Hi Nahla

My team were dealing with a similar IBM i Backup issue recently.  We started by checking out the backup hardware and the backup job and you should do this but we soon found that was fine, instead we would that the system was still in use whilst the backup is running, in our case more and more ODBC jobs where connecting as the week went on.

So, once you have checked our the obvious hardware and backup possibilities, then I would recommend you use the native Performance Data Investigator (PDi) tooling on your IBM i.    PDi is collecting performance data by default and is installed on your system by default, it comes and zero extra cost and is one of the most precious Hidden Gems on the platform.   Use PDi to work out what else is going on your system during the period of the backup.

https://developer.ibm.com/tutorials/ibm-i-performance-data-investigator/

Good Luck,  Brad

------------------------------
Steve Bradshaw Friendly Techie Bloke
------------------------------

Dear Nahla

I assume you use save-while-active.

Are you aware that there were any remote SQL access connection(s) to your system during the data backup period?   You indicated that you ended all application subsystems only which means that remote SQL access from outside of IBM i could still be active and it might lock some objects before you started the data backup process which caused save-while-active to wait for it to finish and close the connection. 

One possible way you can consider trying is to also run the command ENDHOSTSVR SERVER(*DATABASE) ENDACTCNN(*DATABASE)  before running data backup.  But please keep in mind this means any active remote SQL access will be terminated.  When data backup finishes, you run STRHOSTSVR SERVER(*DATABASE) and start all your application subsystems.  If this solves your issue, then you know the cause of your issue and you should notify all parties who use remote SQL access to IBM i database to properly close the connection when finishing the access.  Or ask them to avoid running any remote SQL access during your data backup period. 

If this does not solve you issue, then you should review your performance data during the prolonged data backup period (Tuesday to Friday) compared to the normal period (Monday). PDI (Performance Data Investigator) tool is the best one to use and you should start by looking at the graph named Wait Overview and Wait By Generic Job and see if any substantial wait time exist during the prolonged data backup period or not.   

Hope this help.

------------------------------
SATID SINGKORAPOOM
------------------------------

Hi Nahla,

If you want to find out whether you have connections during your backup, run this statement in ACS or STRSQL (make sure to Change the dates and times to match the backup).  N.B. If you are not using *SQL Naming, change qsys2.history_log_info to qsys2/history_log_info

WITH Connections as (
      SELECT from_user,
         SUBSTR(CAST(message_tokens as VARCHAR(80) CCSID 37), 1,10) as job,
         SUBSTR(CAST(message_tokens as VARCHAR(80) CCSID 37), 64, 15) as ip
      FROM TABLE(qsys2.history_log_info
         (START_TIME => '2021-10-23-22.00.00.000000',
         END_TIME => '2021-10-24-04.30.00.000000'))
      WHERE message_id = 'CPIAD09')
   SELECT ip, from_user as user, job, count(*) as connections
   FROM Connections
   GROUP BY IP, from_user, job

I hope this helps too
Terry

------------------------------
Terry Bartlett
------------------------------