David Martinez wrote:
>
> Hi Alan!
>
> I'm writting an easy visual configuration tool for Heartbeat,
> GTK based in C, and I would like to make you some comments.
>
> I have been reading the Heartbeat code and I think that we should
> modularize
> the configuration related functions like parse_config, dump_config and
> more from
> the main module heartbeat.c, by this way we can share this code for the
> configuration and monitoring tools, and if somebody changes or add some
> variables in the future, it will be trivial to update these tools.
This makes sense to me. I bet you want dump_config to *work* too :-)
[It works for everything but ppp-udp links].
> If you want I would try to make this modularization, but I think this
> functions
> depends on some structs and global variables that you Know better than
> I.
> Also probably I don't have the lastest version of the code.
I can separate those out for you, but you can get the latest version of the
code from the CVS repository.
> Anyway, these are steps I'm following for the configuration assistant:
>
> 1 - Test if the Heartbeat package is installed (What's the best way?)
The *best* way is to install it as part of the heartbeat package :-),
but failing that, check for /etc/ha.d things or /usr/lib/heartbeat things...
> 2 - Serial ports configuration, devices , baud ...
> 3 - UDP conections and network devices configuration. port , ip ...
> 4 - Other parameters like watchdog, keepalive, log files ...
> 5 - Machines in the cluster (I think it should read the /etc/hosts
> file and allow
> the user to select the nodes)
Or possibly allow them to be entered manually as an option.
> 6 - Shared resources for the cluster. Ip-addresses and services to
> start.
Good. I created support for giving text for a GUI config tool in the API for
the resource files. For every file in /etc/ha.d/resource.d, there is an
option called info that will tell you how to configure this correctly. Sadly,
I didn't think about multi-language support. Please look it over.
For example, try /etc/ha.d/resource.d/IPaddr info, and see what I'm talking
about. This output format is subject to change if you have suggestions.
Note that this won't work for those things in /etc/rc.d/init.d.
> 7 - Automaticed start of Heartbeat in the specificated run levels.
Good.
The current code also supports authentication configuration as well.
> Basically these are the steps written in the GettingStarted document.
> By now the program will not be Gnome-compliant but I think to make this
> in
> the future will be trivial.
>
> If you think that part of this mail should be considerated by the people
> of the list, please
> forward it.
>
> Waitting for your considerations, sincerely.
>
> David Martinez
I think this is a good idea. It would also be good for you to look at the
pirhana package and see if there is some code there you can use. I would
suggest sending out your ideas to the linux-ha-dev list, and starting the
discussion there.
Definitely look at pirhana. It configures their heartbeat code along with
performing LVS configuration. It would be nice if we had a tool that worked
together with LVS configuration. You can also check with the LVS group and
see if they have something to start from.
Let me know what you find out, but in the meantime, I've CCed this response to
the linux-ha-dev mailing list.
-- Alan Robertson
alanr@bell-labs.com
>
> Hi Alan!
>
> I'm writting an easy visual configuration tool for Heartbeat,
> GTK based in C, and I would like to make you some comments.
>
> I have been reading the Heartbeat code and I think that we should
> modularize
> the configuration related functions like parse_config, dump_config and
> more from
> the main module heartbeat.c, by this way we can share this code for the
> configuration and monitoring tools, and if somebody changes or add some
> variables in the future, it will be trivial to update these tools.
This makes sense to me. I bet you want dump_config to *work* too :-)
[It works for everything but ppp-udp links].
> If you want I would try to make this modularization, but I think this
> functions
> depends on some structs and global variables that you Know better than
> I.
> Also probably I don't have the lastest version of the code.
I can separate those out for you, but you can get the latest version of the
code from the CVS repository.
> Anyway, these are steps I'm following for the configuration assistant:
>
> 1 - Test if the Heartbeat package is installed (What's the best way?)
The *best* way is to install it as part of the heartbeat package :-),
but failing that, check for /etc/ha.d things or /usr/lib/heartbeat things...
> 2 - Serial ports configuration, devices , baud ...
> 3 - UDP conections and network devices configuration. port , ip ...
> 4 - Other parameters like watchdog, keepalive, log files ...
> 5 - Machines in the cluster (I think it should read the /etc/hosts
> file and allow
> the user to select the nodes)
Or possibly allow them to be entered manually as an option.
> 6 - Shared resources for the cluster. Ip-addresses and services to
> start.
Good. I created support for giving text for a GUI config tool in the API for
the resource files. For every file in /etc/ha.d/resource.d, there is an
option called info that will tell you how to configure this correctly. Sadly,
I didn't think about multi-language support. Please look it over.
For example, try /etc/ha.d/resource.d/IPaddr info, and see what I'm talking
about. This output format is subject to change if you have suggestions.
Note that this won't work for those things in /etc/rc.d/init.d.
> 7 - Automaticed start of Heartbeat in the specificated run levels.
Good.
The current code also supports authentication configuration as well.
> Basically these are the steps written in the GettingStarted document.
> By now the program will not be Gnome-compliant but I think to make this
> in
> the future will be trivial.
>
> If you think that part of this mail should be considerated by the people
> of the list, please
> forward it.
>
> Waitting for your considerations, sincerely.
>
> David Martinez
I think this is a good idea. It would also be good for you to look at the
pirhana package and see if there is some code there you can use. I would
suggest sending out your ideas to the linux-ha-dev list, and starting the
discussion there.
Definitely look at pirhana. It configures their heartbeat code along with
performing LVS configuration. It would be nice if we had a tool that worked
together with LVS configuration. You can also check with the LVS group and
see if they have something to start from.
Let me know what you find out, but in the meantime, I've CCed this response to
the linux-ha-dev mailing list.
-- Alan Robertson
alanr@bell-labs.com