cloudstack-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Wido den Hollander <w...@widodh.nl>
Subject Re: [DISCUSS] Automatic usage reporting / call-home for CloudStack
Date Mon, 01 Dec 2014 13:40:01 GMT


On 12/01/2014 02:15 PM, Sebastien Goasguen wrote:
> 
> On Dec 1, 2014, at 8:08 AM, Wido den Hollander <wido@widodh.nl> wrote:
> 
>> Hello,
>>
>> As a project we currently don't have a lot of insight information on
>> about how CloudStack is being used. Surveys tell us a lot, but not
>> everybody fills in the survey, so we still miss a lot of information.
>>
>> That's why I've written the Usage Reporting functionality for the
>> management server which automatically sends back anonymous information
>> about a CloudStack deployment.
>>
>> It's currently in the 'reporter' branch. [0]
>>
>> By default, every 7 days it generates a JSON document with:
>> - Hosts (Number, version, type, hypervisor)
>> - Clusters (Hypervisor en Management type)
>> - Primary storage (Type and provider)
>> - Zones (Network type and providers)
>> - Instances (Number and types)
>>
>> This report is not complete yet, I'd like to add more information, but
>> that will be Management Server information.
>>
>> The code on how this report is generated is obviously 100% Open Source,
>> so end-users can always exactly see how the information was compiled.
>>
>> I want to discuss this new feature for CloudStack and the possible
>> implications it might have.
>>
>> I'm opting for a opt-out. So every new or upgraded install to 4.6.0
>> (master) will have this enabled. Yes, we have to be very explicit in the
>> Release Notes that this has been added.
>>
>> Why? It's the small price we as a project ask for using CloudStack. We
>> want a little bit of information on how CloudStack is being used so that
>> we can use this to make CloudStack even better.
>>
>> Turning it off is also just one global setting and it will never turn on
>> again.
>>
>> On the server-side there is a Python flask application [1] (found in the
>> reporter directory) which stores all the incoming information in a
>> ElasticSearch database. From there analytics can be gathered on
>> CloudStack deployments.
>>
>> It currently points to http://cs-report.widodh.nl/report which will NOT
>> be the endpoint when this is merged into master.
>>
>> For 'production' I want to have
>> https://report.cloudstack.apache.org/report where all reports are submitted.
>>
>> For every setup a unique ID is determined by hashing the first row in
>> the 'version' table. This is the version + timestamp and that is hashed
>> using SHA256. Using this unique ID we can track changes in deployments
>> and see how they grow or shrink.
>>
>> Technically this wasn't that hard to implement, but the politics
>> surrounding it might be the hardest part.
>>
>> What do other have to say about this? Should there be a VOTE for this
>> feature to come into CloudStack? Opt-in, opt-out?
>>
> 
> We should ask users@ as well.
> 

Yes, that is probably a good idea. First I'd like to have the politics
set before we take it to users.

In the end I'd like to see a VOTE coming up for this feature to go into
CloudStack.

Wido

> I am +1 on this, 
> 
>> Wido
>>
>> [0]:
>> https://git-wip-us.apache.org/repos/asf?p=cloudstack.git;a=shortlog;h=refs/heads/reporter
>> [1]:
>> https://git-wip-us.apache.org/repos/asf?p=cloudstack.git;a=blob;f=reporter/usage-report-collector.py;h=500a4d284b5172fd93acea08f5460cfff5520855;hb=reporter
> 

Mime
View raw message