Re: Full System Save

The term you want to look for is "save while active".

There are a couple of ways to look at this.

The official way is you still close down your payment portals, then you
fire off your save while active, and once it gets a "checkpoint" it will
send a message to a message queue. At this time it starts the physical
save and you can start up your payment portals again without a problem.
An alternative is called a "ragged save while active". In this technique
you tell it not to sync anything and just save it. It may not be in sync.
Meaning that the save times may be a few transactions off. For example
your order header totals may not match the totals from your order lines.
One or the other may have had some transactions hit that are saved at a
slightly different time.

When we used save while active we used the ragged save while active. Not
my choice but the edict from the boss. No outage at all.

Now we do all of our saves from our HA system. Peace and glory throughout
the land. This is the way to go. Users keep banging away on the primary
system while the save runs on the backup system. Totally non disruptive
and the data is in sync. (That is, if you trust the replication
software.)

An Overview of the Save-While-Active Process
http://www-01.ibm.com/support/docview.wss?uid=nas8N1015711

Eliminating your save-outage time
<snip>
Use the following general procedures to eliminate your save-outage time
for particular save operations. These save-while-active procedures do not
require any applications to be ended to perform the save operation.
However, these save-while-active methods do require additional recovery
procedures.
</snip>
https://www.ibm.com/support/knowledgecenter/ssw_ibm_i_73/rzaiu/rzaiurzaiu348.htm
<snip>
4. Objects with a lock conflict still allow checkpoint processing to
complete, and the save operation continues. However, the system does not
save objects with a lock conflict.
</snip>
And that is because they said they said SAVACT(*SYNCLIB). The "ragged"
way uses
<snip>
*SYSDFN
Objects in a library can be saved while they are in use by
another job. Objects in a library may reach checkpoints at
different times and may not be in a consistent state in
relationship to each other.
</snip>
Basically you cross your fingers on a restore.

If you try that *SYNCLIB you'll have to keep a close eye on your joblog
for locks. "If" you do not quiesce all operations while waiting for the
checkpoint.

And since you've mentioned "payment portals" there really is no good sync
when it comes to dealing with "save while active" and IFS objects. And
locks are pretty tight in that realm. Expect to see numerous
CPFA09E - Object in use. Object is...


Rob Berendt