[Linuxha-users] Linuxha.net V2 Design

[Michael Mansour]

Hi Simon,

> 	As of Friday I started designing / prototyping what will eventually
> become V2 of linuxha.net. I already have in mind a series of
> changes/features the newer version will offer, but if you have any
> thoughts on where you think linuxha.net can be improved, or currently
> falls short please let me know!
> 
> 	This does not of course mean that V1 will suddenly stop. V1 will
> continue to be fixed/improved in an incremental manner until at 
> least a year after V2.0.0 actually appears.
> 
> 	And of course at the time the linuxha.net manager software should have
> been "production ready" for some time and will play an important 
> role in cluster infrastructure for V1 and V2 clusters.

Just in regards to this email you sent earlier, I don't know if drbd can do
this, but I would personally like to see at some stage in the future:

* using both nodes in the cluster for clustered services 

I currently have both apache and mysql clusters under linuxha.net. Currently
linuxha.net has the primary/standby node model which I really like, but
(barring the limitations of drbd) it may be an idea in the future to use both
nodes in the cluster, with each running it's own application (one apache and
one mysql for example), and if either server fails, to failover it's
application to the other node.

This would really make good use of the standby node.

* load balancing between both nodes?

I understand there's a limitation to drbd being only mounted on one node at a
time. However reading the drbd website this may change in the future, so if
they do support mounting drbd on both nodes, it would be good for linuxha.net
to be able to build on that and load balance tcp services between nodes using
the dual-mounted drbd device.

* a web-based configurator/installer

I'm personally not a web-based configuration person as I live in shell, but I
think for more widespread adoption of linuxha.net it would be good to have
this on the todo list, to make it easy to not only install and configure, but
create all the necessary files needed for the cluster and app from a web
browser (tick options and bullets and have it create/configure your clconf.xml
and other files).

This may also make it easier when upgrading software for an admin, since new
options can be seen in the browser configuration screen, without having to go
through the clconf.xml.example (and other .example) files and changes/release
notes for the latest updates.

* a test kit (scripts) which can test various failover scenarios once a
cluster has been built

All I mean here is to have something we can run to check the health of the
cluster and possibly recommend actions which can be taken to improve it's
reliability (or fix problems it encounters). This would really speed up the
time in testing the cluster once it's built, as typically I've taken quite a
long time to setup, test and move to production linuxha.net clusters. Anything
that can be done to speed up this process I'd welcome.

Thanks and regards,

Michael.