Hi Andrea,
I will try to answer your questions.
"These applications need to be deployed on average on 10 boxes.
so application A has to be deployed on serv1, serv2....serv10
appl B has to be deployed on serv30, serv31....serv40"
1) For the repeat deployment of a template in several boxes it is not a
problem. That is very easy with SF but what you need to identify is what are
the differences between this boxes if any when starting the application and
decide how you want to start them. For example, you could have a component that
takes a template and a list of hosts and then deploys the same description in
all of them. An example for this is the "spawn" example or what he dynamic
webserver exaple does when the load increases. Check also the array compound.
For a more sophisticated and dynamic system you could use Anubis to perform
some kind of resource management and to discover the status of the
application, the load of the system etc.
Then, I would add some type of monitoring to check if this application are
alive or not. The simplest thing is to do some type of check from the template
that started the application and this could also send you some notification
(email,jabber message) a wait for an admin respond or even try to reboot the
application.
"Each of these application has property files.
We can take an average of 100 property tokens for each application.
The token has to be replaced at the startup of the application in the config
files.
We need a way to manage this token.
We have tokens that are commons to all the environent, other that have
different values for different environment.
so probably we need a DB to store this values and an interface to manage, add
new one...etc...
2)
a) To modify the property files you could write your own component or you could
use the ANT wrapper component and then use the Ant task "Replace" to modify the
values in the property files by using tags for example in the form @[EMAIL
PROTECTED]
b) SmartFrog values be resolved by using "SmartFrog" references. A reference
can be to another description, template, or even a DB or LDAP server. And
therefore it would be possible to get the final values from a DB or similar.
This makes a lot of sense for your case. The other option is that you created
different templates directly in .sf form and you store them in SVN/CVS. This
allow you keep track of changes. I don't know which one could be the best one
for you. I you are planning to offer a customized interface for a user to edit
its values you could have a look at Avalanche that does this. It is in our SVN
server but not yet publicized because we are completing the documentation for
it. This "portal", I think, could be very much in line with what you are trying
to do.
We have 2-3 release per each environment per week.
So we have a very huge amoount of deployment activities and risk to loose token
values.
Because, due to time pressure, guys from operation change values directly on
the boxes.
3) I agree with you. That is never a good idea. The SVN method that I proposed
before forces admins to modify the templates and then the deployment servers
takes over to deploy the changes. You could even automate this.
We want a way to avoid this.
So, from what you said, we need:
- to create templates for each component
- customize this templates in order to get correct values from the token
database.
- use scripts to deploy on enviroment.
In your opinion is it possible to create logical entities like:
- projects
- enviroment
- hosts
- component
- tokens for each component
4) Yes this is exactly what we do all the time. This allows you to break the
problem down very nicely.
If you decide to go ahead with a PoC we could try to help you. We have limited
time but I think your case could be an interesting example for other people.
Steve's book about ANT has a chapter about deployment with SmartFrog that could
be useful to you. Steve can point you where to get it if you are interested and
as I said Avalanche could be very much in line with what you want.
Best regards,
Julio
-----Original Message-----
From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]
Sent: 18 May 2008 20:42
To: Guijarro, Julio
Cc: smartfrog-users
Subject: RE: [Smartfrog-users] info request
Hi Julio,
thank you for your prompt answer.
I'm convinced that SmartFrog with its flexibility could be the right solution
for our big problem.
I can give you an example of the architecture we have and the problems we are
facing.
We have a big portal that consist of 10-15 applications that work toghether.
These applications need to be deployed on average on 10 boxes.
so application A has to be deployed on serv1, serv2....serv10
appl B has to be deployed on serv30, serv31....serv40
Each of these application has property files.
We can take an average of 100 property tokens for each application.
The token has to be replaced at the startup of the application in the config
files.
We need a way to manage this token.
We have tokens that are commons to all the environent, other that have
different values for different environment.
so probably we need a DB to store this values and an interface to manage, add
new one...etc...
We have 2-3 release per each environment per week.
So we have a very huge amoount of deployment activities and risk to loose token
values.
Because, due to time pressure, guys from operation change values directly on
the boxes.
We want a way to avoid this.
So, from what you said, we need:
- to create templates for each component
- customize this templates in order to get correct values from the token
database.
- use scripts to deploy on enviroment.
In your opinion is it possible to create logical entities like:
- projects
- enviroment
- hosts
- component
- tokens for each component
?????
Hope that my explanation is clear enough.
Hope to hear from you soon
have a nice week start
Regards
Andrea
---------- Initial Header -----------
>From : "Guijarro, Julio" [EMAIL PROTECTED]
To : "[EMAIL PROTECTED]" [EMAIL PROTECTED],"smartfrog-users"
[email protected]
Cc :
Date : Thu, 15 May 2008 16:05:34 +0000
Subject : RE: [Smartfrog-users] info request
> Hi Andrea,
>
> Welcome to the list.
>
> One of the main design goals for SmartFrog is to simplify the configuration
> of large systems and to use SmartFrog to configure and manage a system like
> the one you are describing you would have to (in general terms):
>
> 1. Wrap the application software with SmartFrog components that can interpret
> and pass right configuration parameters.
> - In your case, probably, it would be just creating templates for the
> shell scripting component to start/stop your software with the right
> configuration.
> - If your application configuration has to be expressed in the form
> of "property files" then you will need to add a simple component to
> create/modify those files based on the configuration attribute/value pairs
> provided via the SmartFrog language.
>
> 2. Write the configuration templates for those software components, modelling
> their configuration information in the SF language.
>
> 3. Describe the application lifecycle using 1 and 2 and probably some extra
> components to describe how the software is copied to the remoted machines are
> to verify that those machines have the right config and status.
>
> A good example for this type of setup is the "Dynamic Webserver Example". In
> this example we show how to install configure and manage several instances of
> Apache web server. The example shows you how to for example start/stop
> "apached" using the shell scripting component, how to read the number of
> threads used by apached, how to copy and unzip the server in the destination
> machine and so on.
>
> This example is part of the complete SF distribution. For more information
> about how to use it:
>
> Overview doc:
> http://smartfrog.svn.sourceforge.net/viewvc/*checkout*/smartfrog/trunk/core/smartfrog/docs/sfDynamicWebServerExample.pdf
>
> How to set it up:
> http://wiki.smartfrog.org/wiki/display/sf/Dynamic+Web+Server
>
>
> In summary going through your list:
>
> - we need to manage properties in a proper way. We can have property value
> for one configuration and another one for another configuration.
> we want to store information about each property with description and
> documentation
>
> >> Model all this once in SF language and reuse after (templating)
>
> - we need to have automatic deployment
>
> >> Deploy using a SF description
>
> - to deploy on multiple boxes multiple applications
>
> >> No problem. You can either doing by having an SF daemon in those machines
> >> or by using ssh/tftp from a remote one. Your choice.
>
> - to have an inventory system for all the boxes
>
> >> You will need to add some components to do this and it will be specific to
> >> you application. Nonetheless we have standard components that can access
> >> to DBs and so on that can be extended and reused.
>
> - a sort of sanity check
>
> >> Same as previous point.
>
>
> If you need help in trying to understand how everything fits together send us
> a simplified example that could help you to evaluate the feasibility of a
> SmartFrog based solution and we will help you to get it done.
>
> Regards,
>
> Julio Guijarro
>
>
>
> -----Original Message-----
> From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of [EMAIL
> PROTECTED]
> Sent: 15 May 2008 14:03
> To: smartfrog-users
> Subject: [Smartfrog-users] info request
> Sensitivity: Confidential
>
> hi all,
> i'm working for a big telco company.
>
> We stated that we want to optimize our deployment phase, cause at the moment
> we are facing lot of problems.
>
> Currently, we receive the release of the software and then we deploy using a
> perl script.
>
> We have to deploy on severla boxes with different configuration.
> configuration and configuration properties are the biggest problem.
> Operation need waste a lot of time configuring the application.
>
>
> We want to understand if Smart Frog can help us in this challenge.
>
> So our requirements are:
>
> - we need to manage properties in a proper way. We can have property value
> for one configuration and another one for another configuration.
> we want to store information about each property with description and
> documentation
>
> - we need to have automatic deployment
>
> - to deploy on multiple boxes multiple applications
>
> - to have an inventory system for all the boxes
>
> - a sort of sanity check
>
> Can someone explain me at high level point of view how to achieve this
> results???
>
> thanks
> Regards
>
> andrea vavassori
>
>
>
>
> -------------------------------------------------------------------------
> This SF.net email is sponsored by: Microsoft
> Defy all challenges. Microsoft(R) Visual Studio 2008.
> http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/
> _______________________________________________
> Smartfrog-users mailing list
> [email protected]
> https://lists.sourceforge.net/lists/listinfo/smartfrog-users
>
-------------------------------------------------------------------------
This SF.net email is sponsored by: Microsoft
Defy all challenges. Microsoft(R) Visual Studio 2008.
http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/
_______________________________________________
Smartfrog-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/smartfrog-users
