Quantcast

newbie question about agent and console

classic Classic list List threaded Threaded
5 messages Options
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

newbie question about agent and console

b0c1
Hi!

I installed the application to local environment. That's cool, it's work (Easy production setup). Now I want to install only the agent to the hosts, but the documentation said: "All the distributions that were generated during Step 4 now need to be installed on each host. "

So I need install the zookeeper, and the console too? I think I need install only the agent, it's right?
If I installed the application local that's fine but I need to create a different distribution to the hosts? (Because the hosts not run zookeeper, but it's need to the registration...) Or what is the prefered way?

Thanks b0c1

ps: Another question, I want to register Vagrant machine's to my server (I using chef to pull up)
1. There are any cookbook for this?
2. The agent's identify the server with the key. But there are any option to made a name to each agent (if I have 4 Vagrant machine which agent is the right)
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: newbie question about agent and console

frenchyan
Administrator
What is important is that in your meta model, you fully declare where each component go. You should not have "localhost" or 127.0.0.1 when you start splitting across hosts or your system will not work.

So if you have generated your original setup and installed it with "localhost" then you would need to regenerate it with the proper host names and reinstall it (even if on the same node as before).

In regards to your vagrant/agent name questions, the idea is that by default the agent gets "registered" to glu based on the host name of the machine it is installed on (InetAddress.getLocalHost().canonicalHostName by default). If you want to install more than one agent per node, then that will obviously not work since 2 agents will have the same name. I am not too sure what vagrant does, but if each vagrant instance is a new "node" (meaning InetAddress.getLocalHost().canonicalHostName would return a different value even if physically running on the same host), then you don't have to do anything. Otherwise you would have to specify a different name for each agent (in your meta model), which will generate 1 tar file per agent name (this is why it is better to not have to name the agents, because you can then install the SAME tar file everywhere...)

Yan 
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: newbie question about agent and console

b0c1
Thanks the info.

Please correct me. I download glu for my main server (here run the console). I configure the database, the console host, the zookeeper cluster address (now let's configure to localhost), and agent host. Start the zookeeper cluster, install console (not required agent install now)

- What happened if I want to create agent which not configured yet? I define in the meta-model and regenerate the whole distribution?
- In the documentation you using /tmp to store meta info, distrib and so on, but it's deleted after restart. So all generated keys is lost (maybe change in the doc? or it's not required after console install?)
- The install script install the agent locally (with cp) and install the app remotely (scp). Why? Mostly Users install agent from console (main) server with scp. Or you copy the generated distro (by hand) to every agent, and start the agent setup?
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: newbie question about agent and console

frenchyan
Administrator
Thanks the info.

Please correct me. I download glu for my main server (here run the console). I configure the database, the console host, the zookeeper cluster address (now let's configure to localhost), and agent host. Start the zookeeper cluster, install console (not required agent install now)

- What happened if I want to create agent which not configured yet? I define in the meta-model and regenerate the whole distribution?

Yes you would change the meta model and regenerate the distribution.
 
- In the documentation you using /tmp to store meta info, distrib and so on, but it's deleted after restart. So all generated keys is lost (maybe change in the doc? or it's not required after console install?)

This is true. I will emphasize it in the documentation. You should indeed at least keep the keys as once you generate them you don't want to generate new ones (otherwise it means reinstalling every component of glu, which is probably not what you want).
 
- The install script install the agent locally (with cp) and install the app remotely (scp). Why? Mostly Users install agent from console (main) server with scp. Or you copy the generated distro (by hand) to every agent, and start the agent setup?

The install script tries to do the "best" thing based on the host. If localhost, it uses cp. Otherwise it uses scp (because in general you install glu on "remote" nodes). The install script is generated as an example of what you need to do on each host using cp and/or scp. If you want to install with a different method, feel free to do it manually using other techniques (ftp, etc...) but you can use the script as a guide for paths, etc...

Yan
Reply | Threaded
Open this post in threaded view
|  
Report Content as Inappropriate

Re: newbie question about agent and console

b0c1
Ummm. I saw the install-all.sh script and I only see the cp command in the agent... maybe this disturbed me :)

So better if I not do /tmp in the application or store the key's to safe place...
Thanks
Loading...