Re: Anyone in the group running Vision Solutions Mimix 8.1 for HA/DR ?

We used to use Mimix but now use Quick-EDD. Hasn't been that long ago so
my memory is reasonably fresh.

We use HA to go
Production Primary -> Production HA
Development Primary -> Development HA
We do our daily backups on our HA lpars.

Our HA machine is in a data center out of state, a short drive from Larry
Bolhuis.

We kept Mimix updated. They came out with updates on a monthly basis,
pretty much the same day of the month. We put them on promptly unless we
had a planned switch that week.

We switch quarterly, on a weekend. We run production on the other machine
during that weekend. You do find some issues regularly, like NetServer
share names not defined, etc but no real show stoppers.

We have two EMC Data Domain 2500 Virtual Tape Libraries. One in each DC.
They replicate between the two. I have tested restoring data from a
replica. I have done bare metal unload/reloads from these devices. We do
our nightly saves on the HA machine and a quarterly full system save on
all lpars.

We do not do querying on the target machine. Load on the primary is not a
concern.

For OS upgrades we upgrade the target/HA machines first. This stops the
TGTRLS issues. In theory it also gives you another test. But we try to
do that on our sandbox machine first.

For application upgrades all this additional setup has little to no
benefit. For example, it's not like we could run on our HA machine while
upgrading from ERP x.x.x to ERP x.x.y on our primary machine.
I have a tough time envisioning how that would be possible. Most
replication products tend to be one way at a time. Sure Domino supports
two way replication but even that can get into "replication/save
conflicts".
Picture this:
Normal: You make a update to data within a file on the source machine.
The updated data is sent to the target machine and a job posts the update.
All updates to the data on the target machine outside of Mimix are locked
from happening.
Switch: You make a update to data within a file on what is now the new
source machine. The updated data is sent to the target machine and a job
posts the update. All updates to the data on the target machine outside
of Mimix are still locked from happening.
Basically your update goes to a journal receiver. A remote receiver on
the target machine receives the update. A job on the target machine posts
the updates from the receiver.
If one machine is down then the updates stack up until the remote machine
starts processing the updates again.
How would you expect such two way replication to work? Let's say your ERP
update changes the transaction date from 8,0 to a true date field and runs
a conversion. Meanwhile people on the other machine are banging out
transactions using the old versions. How would these HA solutions merge
the two conflicting updates?
If this is the cure you're looking for you're SOL.
This is where you have to just run a conversion on the test library and
plan how long it will be when you do it live and schedule an outage.



Rob Berendt