Controller SPOF

[Cloudy]

Hello All

I've used 3tera in the past. I've evaluated the AppLogic product. One of the issues that I found while testing was that if the controller VM died all the other VMs needed to restart. Is this still the case? If so what are 3tera recommendations on working around the single point of failure ?


Thanks

[PeterNic]

Cloudy,

Yes, this limitation is still in place (stated in the release notes). Resolving it is on the roadmap for '08.

Over the last two years, we have gone through quite a few hardware failures -- servers do fail, and as we grow our footprint, there is a larger and larger number of servers. None of the failures has created (or could create) any serious problems; most of them were automatically recovered within minutes.

Most of our DR and downtime-sensitive customers have grids in two or more different datacenters, because they anyway need to tolerate even complete datacenter failure. With AppLogic in general running this and keeping it simple is, if not easy, at least much easier than in any other way. We also released the first component of our DR suite, MYSQLR and have assisted customers in setting up geographically distributed applications.

I am not saying it is not a problem... it is just less of a problem as most downtime-sensitive projects require DC-level redundancy. If you want to discuss this in more detail, I'll be happy to set up a call with our SE.

Regards,
-- Peter

[Cloudy]

Hello Peter

Thank you for the quick response. Is the ability to put the controller VM on a high redundancy server available ? The usage of the grid would be to power an application and the budget would now allow for a DR solution, which is usually the case with start ups.


Thank you for your time

[PeterNic]

Cloudy,

Sorry - I thought I had responded to this...

Sure you can use a high-redundancy server -- in Clintonian fashion I'll ask you to define a high-redundancy server, though. I am sure that some of our hosting partners may offer this (e.g., an HP box with redundant power supplied, redundant fants, etc.). However, it will become more expensive than two small grids... even two dev grids for under 1k can give you the DR you seem to need.

If you like, you can contact our technical pre-sales to discuss your specific needs (unless you are comfortable posting them here) -- in terms of resources you need to start, downtime tolerance you want to achieve, etc.

Thank you for your interest in AppLogic!

-- Peter

[Karl]

Is there any time scale on the redundancy of the controller VM? Whilst in an ideal world all clients would run copies of their apps on two grids in diverse locations to provide the highest redundancy, that's not always an option and a single grid provides a half way house, although VMs needing restarting if the controller dies, goes someway to defeating that.

[PeterNic]

Karl, this should be available in a major release either later this year or early Q1'09.

Best regards,
-- Peter

[Karl]

I'm beginning to think that you live on these forums :) Cheers for the quick reply on that. Don't suppose there's a road map hiding away anywhere in the docs/wiki that I'm missing?

[PeterNic]

I'm beginning to think that you live on these forums :)
Yes -- my wife and kid come to visit me here often :)


Cheers for the quick reply on that. Don't suppose there's a road map hiding away anywhere in the docs/wiki that I'm missing?

You are welcome. We do share bits and pieces of the roadmap with strategic customer accounts; much easier if you are already a customer. Also, as time for releasing a feature approaches, there is more info posted here and on our blog about it. If you have interest in particular items, please PM or contact pre-sales.

Thanks for your interest!

- Peter

[PeterNic]

Hello All
I've used 3tera in the past. I've evaluated the AppLogic product. One of the issues that I found while testing was that if the controller VM died all the other VMs needed to restart. Is this still the case? If so what are 3tera recommendations on working around the single point of failure ?
Thanks

Cloudy, Karl,

The controller restart and HA has been implemented in 2.7 beta and has been working quite well for a few months. It is going to be included in the 2.7 production release which is coming up in a couple of weeks.

There are quite a few other improvements for uptime, including diagnostics, self-tests, early storage failure warnings, etc. See http://doc.3tera.com/AppLogic27/ as well as http://doc.3tera.com/AppLogic27/High-Availability.html specifically on how it works now.

I am hunting down the threads where this issue has been discussed, hoping to respond to everyone who asked at some point or another.

Best regards,
-- Peter