LightCloud Manager must be run on each server?

[Mitchell Hashimoto]

Hi,

Just taking a look through your examples and the manager's code and it
runs all commands locally on the machine so does each server which
runs TT nodes need to have its own manager? Then to start up the
servers I have to go to each server one-by-one and start them up? Or
does the python manager support some sort of SSH to remotely start up
the servers?

Thanks,
Mitchell Hashimoto

[amix]

Hi Mitchell

LightCloud manager controls Tokyo Tyrant servers and you only need to
have one manager running pr. server you are running Tokyo Tyrant's on.
It does not support external access - so you need to SSH into those
servers to control - but it would be great if someone extended it with
a XMLRPC interface ;) [personally I don't need remote control].

Regards,
Amir

2009/3/5 Mitchell Hashimoto <xmi...@gmail.com>:

[Khánh Duy]

I think the manager is not very useful without remote control (except
for, like, all status)
I might modify the manager.py and add ssh to the command string.

I really want to hear amix explain more on this:

you only need to have one manager running pr. server you are running
Tokyo Tyrant's on.

BTW, I'm having a test system with 4 machines, each with 1 lookup and
1 storage node.

[Khánh Duy]

Please make clear about the usage of lightcloud_manager with multi-
server systems.
I need to start/stop/status all the servers from 1 terminal. Is that
possible with the current lightcloud manager?

[lhl]

Amir, I'm also confused about this. Based on my understanding, if you
had a setup w/ master-master replication on three machines machines,
you would need to have three copies of the config file and run the
manager on each system to start? I noticed that it tries (and fails)
to start ttservers on remote hosts?

The process for adding additional nodes is to modify each config file,
then start the manager on the new node, and then modify the client to
point to the new nodes?

[amix]

2009/4/17 lhl <rand...@gmail.com>:

>
> Amir, I'm also confused about this.  Based on my understanding, if you
> had a setup w/ master-master replication on three machines machines,
> you would need to have three copies of the config file and run the
> manager on each system to start?  I noticed that it tries (and fails)
> to start ttservers on remote hosts?

> The process for adding additional nodes is to modify each config file,
> then start the manager on the new node, and then modify the client to
> point to the new nodes?

This is correct. You could automatize this process thought via for
example Fabric ( http://www.nongnu.org/fab/ ).

Regards,
Amir ( http://amix.dk/ )

[Phoenix Sol]

Fabric  =  Lightcloud  =  'a nice surprise'  =  'something I thought I was going to have to implement for myself'

Thanks a heap, Amir!

( Hell, thank (all of you nice hackers out there) for Open Source! )