Page 1 of 1

[RPG]Mirroring LongRange system to HA

Posted: Thu Jul 09, 2020 5:31 pm
by MegumiSawada
Hi,

There're 2 IBM i boxes and one is hot standby for disaster recovery.
I would like to use High Availability tool for keeping these IBM i systems(including LongRange) in sync.

I would like to confirm that my understanding below is correct:
- I need to install the same version of aXes and LongRange on both machines
- I need to get and apply license for both machines(separately)
- I can mirror any objects between the primary and your secondary boxes including aXes libraries and root directories EXCEPT LongRange license(Certificate_NNNNNNNNN.dat)

If my understanding is wrong, or is there any objects which should be excluded from mirroring objects, please let me know.

Thank you in adcvance.
Best Regards,
Megumi Sawada

Re: [RPG]Mirroring LongRange system to HA

Posted: Mon Jul 13, 2020 5:37 pm
by MegumiSawada
If you are wanting to do a HA, you are required to install the same version of aXes on the secondary machine and get licenses for this.

Then you can mirror any other objects between the primary and your secondary boxes including aXes & LongRange libraries and root directories

Things that can be excluded are any logs or trace files.
- /axes_root/logs
- /axesjsm_root/jsm/instance/trace
- /axes_root/LongRange/certificates
- /axes_root/LongRange/tracing

Re: [RPG]Mirroring LongRange system to HA

Posted: Tue Jul 14, 2020 1:11 pm
by jmrapin
There is a vast difference between hot stabndby and HA.

LongRange use aXes in the background.
True HA is not possible with aXes.

The following scenario may not work.
True HA means that a user can work on IBMi1 and if IBMi1 goes down IBMi2 will take over seamlessly. For this you correctly said that you need to mirror all objects. But that's not enough. Let's assume there is a load balancer that detects when either IBMi goes down.

In the case of aXes, we keep a record of all transactions and the order they are made. There is a new signature being sent for all transaction. Therefore if a user is working on IBMi1 and it suddenly goes down. The load balancer will see this and will route the requests to IBMi2. However IBMi2 does not know about the current work so the login screen will be presented. and the current session will have to be restarted.

So for Longrange, this might be the same and users might loose their current work they are working on when IBMi1 goes down.