Diff for "LEP/ApportOnStable"

Not logged in - Log In / Register

Differences between revisions 1 and 2
Revision 1 as of 2011-03-28 10:51:37
Size: 1738
Editor: jml
Comment:
Revision 2 as of 2011-03-28 11:51:19
Size: 2067
Editor: lifeless
Comment: notes
Deletions are marked like this. Additions are marked like this.
Line 28: Line 28:
   -- are flavours of MapReduce sufficient to meet this constraint? If not, there is a conceptual problem here as 'big data' and 'ad hoc' are not friends. [All the big data db's advise design-for-your-queries]. -- RobertCollins
Line 55: Line 56:

RobertCollins:
 * Learn from apport - its privacy behaviour will likely need to be ported over.

Apport on stable

Currently, Ubuntu collects crash data for applications that are being run during the alpha and beta releases of Ubuntu. It would be nice to be collecting crash data all of the time. The problem is that at the moment, crashes are recorded as bugs on Launchpad. They aren't really bugs though, they are crashes. If we instead had some kind of big data system that stored crashes as crashes and allowed people to analyze them on demand, we would provide a much more useful service.

Contact: Jonathan Lange
On Launchpad: Link to a blueprint, milestone or (best) a bug tag search across launchpad-project

As an Ubuntu developer
I want to be able to learn about crashes on stable releases of Ubuntu
so that I can prioritize work for the next release

Rationale

XXX

Stakeholders

  • Mark Shuttleworth, 2011-03-25
  • Robert Collins, ???
  • Kate Stewart, ???

Constraints and Requirements

Must

  • Scale to tens of millions of users
  • Provide ways to do ad hoc queries
    • -- are flavours of MapReduce sufficient to meet this constraint? If not, there is a conceptual problem here as 'big data' and 'ad hoc' are not friends. [All the big data db's advise design-for-your-queries]. -- RobertCollins

Nice to have

  • A way of turning many crashes / a kind of crash into a bug

Must not

  • Notify Ubuntu developers of incoming crashes

Out of scope

  • The system need not keep all data for all time

Success

How will we know when we are done?

How will we measure how well we have done?

Thoughts?

  • Use Cassandra or something similar
  • Need to separate crashes from bugs in Launchpad data model
  • Don't need to start with a UI
  • Find platform interest
  • See how open we can get it, can we just run a public ro database?
  • Privacy is a concern

RobertCollins:

  • Learn from apport - its privacy behaviour will likely need to be ported over.

LEP/ApportOnStable (last edited 2011-03-28 11:51:19 by lifeless)