oVirt.org Access Needed

Brian Proffitt bproffit at redhat.com
Thu Jan 23 21:08:31 UTC 2014


The data Bitergia is collecting for oVirt, Gluster, and RDO is part of a dashboard project that these teams have been working on for some time now, and the end results are finally ready to post to the oVirt community to get their feedback. What you found in [1] is something that we want to host locally on ovirt.org, using a daily cron job that will pull the cached data from Bitergia's git repo found at [2]. 

This has already been done by Rich Bowen at RDO, and the results are at [3].

There is no urgency for this task. I did talk to Karsten about it a while back, but my mention of it on the [infra] list is really my first effort to get this going. 

As you requested, the guidelines sent to me by Rich were:

1) Checkout the [Bitergia] git repo somewhere - anywhere - on the web server. 
Ensure that the directory is permission +rx so that Apache can descend 
into the directory and read the files.

2) git pull every morning with a cron job:

0 4 * * * cd /var/redhat-rdo-dashboard/browser && git pull origin master

(Not sure about the time - if you can tell about what time every day 
they push the updates, that would make better timing. I haven't done 
that yet, so I'm probably a day behind. I should check.)

3) Point Apache at it:

         Alias /stats /var/redhat-rdo-dashboard/browser

         <Directory /var/redhat-rdo-dashboard>
             Options FollowSymLinks

             Order allow,deny
             Allow from all

             DirectoryIndex index.html
         </Directory>

4) Restart httpd and you should be golden. That'll crate a URI of /stats 
for that content.

Let me know if you have any additional questions.

Peace,
Brian

[1] http://bitergia.com/projects/redhat-ovirt-dashboard/browser/
[2] https://github.com/Bitergia/redhat-ovirt-dashboard
[3] http://openstack.redhat.com/stats/




----- Original Message -----
From: "R P Herrold" <herrold at owlriver.com>
To: "Brian Proffitt" <bproffit at redhat.com>
Cc: "oVirt infrastructure ML" <infra at ovirt.org>
Sent: Thursday, January 23, 2014 3:30:18 PM
Subject: Re: oVirt.org Access Needed

On Thu, 23 Jan 2014, Brian Proffitt wrote:

> Thanks to all for the assist. Rich just sent me the how to, 
> and apparently, I need to ssh into the actual server to pull 
> the git repo of Bitergia's data into the web server, set up 
> the alias in the conf file, and add a cron job to pull the 
> git data down daily.

Could you please forward those instructions into the infra ML, 
so the list members can see how it was set up?

and so to some questions for clarification: Is 'bitergia' and 
its sub-parts packaged into a form that has landed in Fedora, 
or ... where ?  I found the demo instance pointed to to be 
very sluggish and loady on my local browser, but did not run 
down why yet with a local install

I spent some time looking for a way to retrieve the sources to 
do a local setup, and was not able to find them.  Are they 
under a FOSS license?  

Is that linked instance pulling real time stats from live 
servers or from cached details?  If the former, infra probably 
need to get an understanding as to the load effects, as the 
ovirt infrastructure lacks spare capacity in terms of memory, 
and in some cases in terms of network bandwidth.  No surprises 
there as it has been reported in infra meetings, and in the 
Wednesday 'sync' but ...
 
> So, is that do-able? Or would it be easier for someone on 
> the team to set this up?

'do-able' and 'done right' probably are different here.  The 
demo instance shows it _can_ be done, but ...

/me looks for a soapbox and puts on an infra 'hat':

	unpackaged tools are 'magical'  magical is a problem

Unpackaged tools are un-vetted in a traceable manner as to 
License.  Unpackaged tools in a remote VCS can simply 
disappear, be invisibly compromised, or go through an API 
change, or otherwise become NON re-deployable in the future.  
A start-up vendor can close its doors and disappear, re-license, 
take down archives, ... .  Entropy happens all the time

The discipline of packaging prevents many of these problems 
from gaining a toe-hold, by forcing retrieval of a version,  
which may be checked against published md5sums (sha, 
whatever); gets a review (by human eyes); and gets replication 
of the build process by a non-human auto-builder.  If there is 
a good 'make test', it also has a sample set of configs to 
read and confirm function of ...

Ones which ** require ** a manual [woops -- magical ;) ] 
content deployment from git from instructions conveyed in a 
non-public channel (or: unrolling a tarball, running some tool 
which untraceably solves dependencies, such as 'cpan' or 'npm' 
to get a point in time image which may be broken tomorrow when 
one goes to re-deploy it) are broken.  

It may be pulling in encumbered matter (eg: a patent problem 
like the old: 'gif', or 'rar'). An undocumented manual 
configuration / setup process is inherently fragile

New non-packaged matter needs explain the path it will follow 
to move to being packaged.  In the interim, it needs the rigor 
of being documented software 'engineering'.  One road to that 
documentation of process is for it to be done via a VCS 
checking, and a puppet CO in deployment, just like management 
of configurations, and re-doable and relocatable via puppet

Also, I do not find that this 'bitergia' facility has a 
tracking bug asking that it be installed [1] [2]

If there is an urgency to attain it, today is the first it has 
been mentioned publicly

The approach of 'undocumented' except via back channel 
communication (digging through IRC logs, digging through email 
archives, whatever) is a broken one.  The result can turn into 
a subject to a single point of outage when the manual maker of 
the tool is off line or unavailable otherwise


Turning to process for 'infra', it seems to me a person 
proposing something should be able to point to a solution 
which has it 'solved' in demonstration with a puppet recipe 
that is checked into a testing branch of the infra git, which 
recipes handle building that a testing instance, and then 
managing its configuration (and so, when re-pointed to 'live', 
takes it live)

-- Russ herrold

[1] https://fedorahosted.org/ovirt/report/1?page=1&asc=0&sort=created
[2] 
http://bitergia.com/projects/redhat-ovirt-dashboard/browser/



More information about the Infra mailing list