reverting commit

Elias Ross genman at noderunner.net
Tue Mar 26 17:28:18 UTC 2013


I do have a couple of problems with the current setup and the new setup.

Current setup:
* Settings file are in a user's home directory ('rhq') whereas
typically they should go into /etc on Linux.
* Settings are in XML which is hard to update using scripts.
* Settings that are per host (agent key) and standard across all
installs are in one file.
* It is very difficult to change the server hostname in an automated
way (i.e. using Puppet). I have had to change many agent hosts (> 50)
to connect to a new server at a new address. I resorted to changing
XML using Augeas which is non-trivial to do.

New setup: (as proposed):
* Unclear how updates are done. I use RPM to install the agent, since
I am using Puppet.
* Unclear what problems are being fixed. Bugzilla doesn't reveal the
issues to me; I assume they are protected.
* It's a bad idea to store these sorts of settings in an application's
working directory.

Things I'd like to see:
* Agent key is stored in a separate file. The reason is that across
all instances I want to use the same properties file, and this is the
only setting that is unique per host.
* Settings are kept in plain text, like a properties file. This makes
it easy to update using scripts.
* Settings are in /etc. On Windows, I don't know what to suggest.
* RPM for the Agent. Okay, separate issue. But makes it easier using
Puppet to do the initial install via RPM. RPM should have an option
for 'rhq' user install.

On Tue, Mar 26, 2013 at 9:28 AM, John Mazzitelli <mazz at redhat.com> wrote:
> [Ian - feel free to commit this back in a branch (not master) and create a BZ for this. We'll cherry pick in master at the appropriate time.]
>
> Its been requested that we revert commit df18b71df2d5114d9edc8a26ad42e0d7a9036eb2:
>
>     "do not pass not-yet-fully-initialized "this" reference to GroupOperationHistory.addResourceOperationHistory() in ResourceOperationHistory's constructor, as the GWT compiler fails when it sees this (at least in my OS X 10.7.5, Oracle JDK 1.7.0_17 environment)"
>
> We are reverting for a couple reasons:
>
>     * This broke a jenkins build - an API change has been introduced and it therefore fails the API check job. (I think its due to a change in a public domain object's constructor). Keep in mind that for awhile now we have a jenkins job that runs to make sure there are no backward compatible changes to public APIs - if it detects one, the build will fail. In this case, it appears a public domain object's constructor was changed thus breaking that API check.
>
>     * Also, There is no BZ for this - we don't know what this addressed. Please make sure BZs are written before committing bug fixes so they can be tracked appropriately.
> _______________________________________________
> rhq-devel mailing list
> rhq-devel at lists.fedorahosted.org
> https://lists.fedorahosted.org/mailman/listinfo/rhq-devel


More information about the rhq-devel mailing list