Re: How long does a PWRDWNSYS restart(*no) actually take?

Le 28/07/2017 à 16:17, Jim Hawkins a écrit :

I have been tasked to write a program to monitor the UPS on our IBM power
system. I found a good base program-so that is all good. We have a generator
that is supposed to kick in within the 1st minute of a power failure. My
question is more of "what if the generator fails?" From my research, it
appears that the ups on the system will last for about 20 minutes.

We do an IPL each week (pwrdwnsys *immed restart(*yes)), (it is what it is,
I'm not rocking the boat on this one, so save your breath). What I see in
the log is entries indicating: this ended, that ended, etc. for about a
minute and a half, then nothing for 32 minutes when I see "Unattended IPL in
progress".

If the generator fails, I want to power down the system, but I am not sure
how much time I have to wait. The last time we did a pwrdwnsys restart(*no)
was 3 years ago when we had to replace the cache battery-and I don't recall
that far back (lol).

I want the system to power down on my terms, not the terms of the UPS. So,
good folks, how much time do I really have to power down, before my UPS runs
out of batteries? (Yes, I will allocate a little extra for the unexpected
and the fact the batteries are 7+ years old.) Just simply, at what point is
the system powered off?

I cannot say how much time it would take to shutdown, it really depends on the circumstances.
However, if I had to work on the same topic, I would investigate two ways in order to try to reduce the shutdown duration.

In the program that you will write, I would try to end as much as possible running jobs with their normal stop procedure. You may have to re-write some of them which includes DLYJOB just in case of. Once done, I would run CHGASPACT *FRCWRT command to initiate flushing memory content to disk without waiting the end of your jobs. In any case, the system will do it for your before shutting down. So, the less it has to write because it is already done, the shorter this step that you cannot control will be (not sure if that sentence is using a proper english, sorry for that).
Second topic, I would decrease QPWRDWNLMT system value which is by default 900 seconds. It means that the system will wait up to 900 seconds before really shutt--ing down, if he is not able to properly end the job. This does not occur often, but it might happen with some java calls, and I have seen it in case of communications troubles with tape devices on SAN.

When I have to shutdown a partition and if I have the opportunity to do it from the console. I always do what I wrote (except decreasing QPWRDWNLMT because I can control if jobs are properly ended):

1-end the jobs properly with their usual stop procedure.
2-run an endsbs *all in a let's say 120 seconds controlled way
3-run CHGASPACT *FRCWRT command
4-wait for the restricted state to complete
5-monitor activity with WRKSYSACT *IO (you will see some tasks, responsible to flush memory content, initiated by CHGASPACT command
6-wait for activity close to zero
7-initiate PWRDWNSYS command in a 15 seconds controlled way

(I will be away from my email from noon today until next week)



TIA


Jim Hawkins

Programmer Analyst

Interkal LLC

Kalamazoo, MI