RE: Leaving IBM i on vs. sleep mode and fire it up now and then

Since the steps needed for a proper power down are so critical, why hasn't IBM coded these steps as part of the PWRDWNSYS?

Paul

-----Original Message-----
From: MIDRANGE-L [mailto:midrange-l-bounces@xxxxxxxxxxxx] On Behalf Of Rob Berendt
Sent: Tuesday, January 17, 2017 7:12 AM
To: Midrange Systems Technical Discussion
Subject: RE: Leaving IBM i on vs. sleep mode and fire it up now and then

That link is nice. It specifically says this about using the white
button:
<snip>
Caution: Do not use this function on a system that has a low or failed write cache battery, as thousands of pages of disk storage can be lost.
This type of IPL will be marked abnormal and will incur significant IPL recovery.
</snip>

The suggested steps are the following:
- Quiesce your user applications. Follow any suggestions from your software vendors.
- ENDTCPSVR *MGTC
- ENDTCPSVR *CIMOM
- ENDTCP
- ENDSBS *ALL *CNTRLD DELAY(120)
- PWRDWNSYS OPTION(*CNTRLD) DELAY(120)
- DSPJOBLOG OUTPUT(*PRINT)
- SIGNOFF *LIST

Let me explain. If you do not end tcp first the ENDSBS will NOT kill all of the tasks starting with the last couple of releases. This is documented. This can cause issues for running tasks while you "thought"
you were in a restricted state.

Failing to end Management Central and CIMOM prior to ending TCP can cause corruption. Been there, done that, have the scars.

When you use *CNTRLD there are flags set on the system and even your user written applications can check for these and end in an orderly fashion.
Even RPG supports %shtdn: dou F3 or %shtdn Granted not all applications code to these flags but allow those which do the opportunity to use that time. After all some poor slob had to write that code, respect their effort.

Yes you can enter a few quick commands after entering PWRDWNSYS; if you do it controlled.

If you've already brought the system down into restricted state, and then issue a PWRDWNSYS *CNTRLD then doing the SIGNOFF will kill the last remaining job and cut the remaining delay off of your outage.

Either 7.3 or 7.2 added ABNENDDLY to the ENDSBS command. If you have a job which will not end without using ENDJOBABN then the ABNENDDLY parameter will automatically issue this after the requisite 10 minutes (that's a minimum, you can override that with a longer delay on this command).

I've never messed with the ENDSBSOPT parameter on both ENDSBS and PWRDWNSYS.
I kind of like the CONFIRM parameter on the PWRDWNSYS command.



Rob Berendt