Client application connecting to wrong system during a HA test.

[rob@xxxxxxxxx]

Resending because I forgot subject line...



When performing a Mimix switch back from our backup machine to our primary 

machine I noticed this message in QSYSOPR:

Date sent  . . . . . . :   08/15/09      Time sent  . . . . . . : 20:29:25
Message . . . . :   Error message CPF4128 appeared during OPEN for file 
  CKACRAMT (C S D F)... program TIMEKEEPER/CLK976...

It was from job:
QSQSRVR 
  QUSER 
  407260

DSPMSGD CPF4128 shows "Not able to allocate objects".  This makes sense 
because Mimix will put a lock on all the objects under replication to make 

sure that you only update them from the current production machine, which 
at this time was our backup machine.  My question was why was it trying to 

access this machine?

That QSQSRVR job didn't start until 8/15 around 20:02.  When we do our 
switch we change IP Addresses and all so anything accessing by IP address 
should have gone to the backup machine.

Looking at the joblog for it I see nothing in that time range:
08/15/09  20:02:38 - Job 407260/QUSER/QSQSRVR started on 08/15/09 at 20:02
very next message:
08/16/09  18:14:55.940375 - ACGDTA for 407260/QUSER/QSQSRVR not journaled

Later on I do see stuff like:
08/16/09  18:16 - User Profile = QNOTES
But I don't think any Notes based application messes with Kronos data. 
Time clocks cache MAC address perhaps? 

Perhaps the time clocks are configured by someone who used NETSTAT *IFC 
and grabbed the wrong IP address (like one of the Domino IP addresses) 
instead of the interface for traditional i applications?

So, since we allow our Domino servers to run on the original primary 
machine, but they were bounced after the mimix switch and therefore 
shouldn't have some route cached or whatever I'm trying to express, why 
would this be happening?  Heck, we've even performed an IPL after 
switching IP addresses and before starting the Domino servers.

If it's some other client application then I would think that should 
reroute over (but obviously something funny is going on).  5250 sessions 
seem to blank out during the switch and appear on the new machine (if you 
have autoconnect and stuff set up).  Maybe my earlier hypothesis about the 

wrong IP address being used by the timeclocks is correct?

Timeline:
Mimix switch (and when IP interface got disabled on primary machine and 
activated on backup machine) roughly 19:30 on 8/14.
IPL finished 8/15 11:32
Domino servers started at 8/15 11:54.
Message appears 8/15 at 20:29

DSPLOG PERIOD((1100 081509)) job(407260/quser/qsqsrvr)
Job 407260/QUSER/QSQSRVR started on 08/15/09 at 20:02:38 in subsystem 
QSYSWRK
Error message CPF4128 appeared during OPEN for file CKACRAMT (C S D F). 
Error message CPF4128 appeared during OPEN for file CKACRAMT (C S D F). 
CEE9901 received by CLK976NOTE at 263. (C D I R) 
CEE9901 received by CLK976NOTE at 263. (C D I R) 
Job 407260/QUSER/QSQSRVR ended on 08/16/09 at 19:33:28; .347 seconds used; 

en

If it was the timeclocks how do I determine which of many, it might have 
been?  The better to inform the person in charge of the timeclocks which 
needs updating?

Nothing in the DSPJOBLOG as you've seen alluding to IP address.
Nothing in DSPLOG about a connection from a particular IP address.
No active comm traces on that line.

Darn things are probably DHCP and picked a new IP address by now anyway.

Sorry if the message rambles I was thinking as I was typing and bounced 
back around as I developed hypothesis.

Rob Berendt