When powering off an lpar of IBM i you really should end TCP/IP before
performing ENDSBS *ALL. Otherwise bad things can happen and even if you
get the message "system ended to restricted state", you aren't. This
ain't OS/400.
<snip>
Recommendation:
ENDTCPSVR *ALL
MONMSG MSGID(TCP0000)
ENDHOSTSVR SERVER(*ALL) ENDACTCNN(*DATABASE *FILE)
MONMSG MSGID(TCP0000)
DLYJOB DLY(300)
ENDTCP
MONMSG MSGID(TCP0000)
DLYJOB DLY(300)
Possible problems by not following recommendation are listed below:
o LDAP not restarting properly or losing data due to the server not being
ended before subsystems.
o Netserver not restarting due to tasks not ending properly without an end
of the server prior to the ending of subsystems.
o Document New, Receiving CWBLM0017 RC2=28 Error When Trying to Start
PC5250 Session:
o Document New, Telnet Fails after System Save (GO SAVE 21): Database
'DCF Technotes (IBM i)', View 'Products', Document 'Telnet Fails after
System Save (GO SAVE 21)'
o Document N1010546, Management Central Failing to Start and Receiving
Message CWBCO1008 RC 8007:
</snip>
http://www-01.ibm.com/support/docview.wss?uid=nas8N1018598
While ENDTCP kind of does an ENDTCPSVR also, there are serious timing
issues (noted above) by not running the ENDTCPSVR first. I have been
bitten by these.
After using the PWRDWNSYS command,
and verifying, via the HMC, that the lpar is shut down,
then you should power off the system from the HMC. I do not mean power
off the HMC. Shutting down the lpars does not necessarily mean the rack
itself is powered off. An exception to that might be if you flag one of
the lpars as power controlling, if so, make sure you power off that lpar
last. Anyway, verify via the HMC that the system unit is powered off.
Then, if you also want to reduce the load on the UPS by powering off your
HMC you may do so.
When you start up the HMC comes first. Then use it to start the rack. Not
the power button on the rack.
I have done data center moves, etc and I do not touch the power button on
the rack.
Rob Berendt
midrange.com
