Servoy and failover

[lee]

I came across this thread mentioning failover and noticed that it's from 2003! With no other hits on the forum mentioning failover, and support planned for 2004, what is the current status of this feature? It would be extremely useful in our situation to have support for this built into Servoy. Would there be any licencing issues with this, or would it not be a problem because of the concurrent connections model?

[Jan Aleman]

Due to low demand the feature has been moved ahead. The architecture is ready for it though so don't let it stop you from starting development. At the moment you could already do failover using hardware failover box(es) in front of the servers.

[maria.kyselova]

Jan, with the current Servoy 6.1 architecture, provided the clustered servers are used, what kind of experience would the user be getting in case of a failover event?

Cheers,
Maria

[Andrei Costescu]

Please start a new thread for a new question the next time. Reopening very old threads with out of context questions makes the forum unorganized.

If you set-up a cluster with load-balancing in 6.1.x, 6.0.x, 5.x:
- as soon as the load balancer realizes that one of the servers is down, it will start redirecting new users to one of the other servers;
- a failure page on the load balancer could be set up to show a message for users connected to that server - and redirect to the login form after a number of seconds;
- the user needs to start the solution again / re-login (the working session on the server that stopped working will not be recovered).

[maria.kyselova]

Please start a new thread for a new question the next time. Reopening very old threads with out of context questions makes the forum unorganized.

If you set-up a cluster with load-balancing in 6.1.x, 6.0.x, 5.x:
- as soon as the load balancer realizes that one of the servers is down, it will start redirecting new users to one of the other servers;
- a failure page on the load balancer could be set up to show a message for users connected to that server - and redirect to the login form after a number of seconds;
- the user needs to start the solution again / re-login (the working session on the server that stopped working will not be recovered).

Thanks, Andrei.
So the unsaved data on the client will be lost in case one of the servers in the cluster goes down?
A redundant server will still be available but it doesn't solve the problem of losing the data.

Are there other any strategies to avoid the loss of data and keep high availability of a Servoy application except setting up the clusters?

Cheers,
Maria

P.S. Sorry about the mess with the forum threads. I thought spawning new threads while there are existing ones with the same name was what made the forum admins angry

[Jan Aleman]

You want unsaved data to be saved amidst a crash? That sounds like a bad idea. See Acid in the description of ACID

http://en.wikipedia.org/wiki/ACID

[Andrei Costescu]

Yes, you are correct about unsaved data and availability.
If we want to do more about unsaved data in this (normally rare) situation it would become a matter of balancing performance vs. redundancy. Maybe in the future.

But Jan is right, what if the server went down while partially executing a script. Then unsaved data might be uncoherent - would it be really useful to recover it?

[maria.kyselova]

Yes, you are correct about unsaved data and availability.
If we want to do more about unsaved data in this (normally rare) situation it would become a matter of balancing performance vs. redundancy. Maybe in the future.

But Jan is right, what if the server went down while partially executing a script. Then unsaved data might be uncoherent - would it be really useful to recover it?

Maybe not.
I'm fishing for options, Andrei. There may be something out there I'm not aware of.
Thanks for your feedback, guys!

Cheers,
Maria