incubator-ambari-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Elazar Leibovich <elaz...@gmail.com>
Subject Re: Support OS without packaging system
Date Mon, 08 Jul 2013 12:49:14 GMT
Hi Siddharth,

Thanks for your answer!

Can you refer me to a documentation or to a similar piece of code
examplifying the usage of puppet Resource providers? I haven't found
something like that in the Ambari Design link you provided.

Thanks again!


On Sun, Jul 7, 2013 at 8:40 PM, Siddharth Wagle <swagle@hortonworks.com>wrote:

> Hi Elazar,
>
> Thank you for looking into Ambari. Following is the link to the Ambari
> wiki with information regarding Ambari architecture that you might find
> useful,
>
> https://cwiki.apache.org/confluence/display/AMBARI/Ambari+Design
>
> Amabri agents that heartebat with the ambari server are written in python
> and delegate the responsibility of installing packages to puppet scripts.
> The agent architecture utilizes puppet in a sandboxed fashion (limited
> context and responsibility), in order to enable the package management to
> be mutable.
> On a side not, it is actually possible to write a custom Provider
> implementation for puppet and hook it up with puppet Resource provider, I
> have not looked into the details of doing this just know that it is
> supported.
>
> Best Regards,
> Sid
>
>
> On Sat, Jul 6, 2013 at 5:11 PM, Elazar Leibovich <elazarl@gmail.com>wrote:
>
>> Forwarding to ambari-user, since I suspect ambari-dev is primarily used
>> for JIRA.
>>
>>
>> ---------- Forwarded message ----------
>> From: Elazar Leibovich <elazarl@gmail.com>
>> Date: Fri, Jul 5, 2013 at 4:11 PM
>> Subject: Support OS without packaging system
>> To: ambari-dev@incubator.apache.org
>>
>>
>> Hi,
>> I'm currently consider deploying Ambari on home brew hardened Operating
>> Systems which does not feature package managements. You cannot use anything
>> like RPM, deb packages etc.
>>
>> The way I was thinking of handling it is:
>>
>> 1. compile all the required package to a certain shared directory:
>>
>> //fileserver/opt/{hbase,hadoop,oozie}
>>
>> 2. Define fictive package manager for Ambari that would simply copy the
>> files to a node instead of triggering actual installtion script.
>>
>> Ambari will then act as usual and will try to install those packages
>> using my fictive package manager.
>>
>> If the base OS installation has all the dependencies, Ambari should still
>> work.
>>
>> 1. Can that work? I'm don't know too much about Ambari's architecture,
>> besides a few slides decks I found.
>> 2. Is that interesting as a JIRA? I think that this flexibility (work on
>> any linux system at all, if you can ./configure && make hadoop on it -
>> ambari will take care if that) is beneficial for more people than just
>> myself. If it is, I'd like to coordinate my efforts with the project, and
>> to submit the code eventually.
>> Note that this is a change with relatively low risk, since I'm just
>> adding a support for different system, and not changing the usual flow at
>> all.
>> 3. If everything will work correctly, there's at least one big
>> organization that would be using Ambari, I think it'll be good for the
>> project.
>>
>> Thanks,
>> Elazar Leibovich
>>
>>
>

Mime
View raw message