CrashKill/WorkWeek2012/CrashRetrospective

• Media:CrashRetrospective.pdf

Things that we are doing well

• Communications with Socorro team
• Release - good job nominating issues and tracking
• Good at release to release comparison - crashes per 100 ADU
• Rapid release - pretty good at fixing code crash regressions introduced in a release cycle
• Mitigate out of memory crashes
• Outreach to partners to get them engaged in fixing problems - Kev
• Better tools for users who are crashing a lot
• Good additions for mobile
  • URLS
  • Skiplists
  • Top crash reports
• Socorro links to bugs
• Signature summary
• Fixed lots of small things in Socorro
• Not losing crash report data
• Good at logging top crashes

Things we could be doing better

• Getting developers assigned to top issues in particular lower volume crashes
• Communications to the lists/blog
• Less good at working lower priority crashes - long tail
• Engineering community not as connected with crash process
• No clear way to communicate good metrics - how we are doing
• Triage - lots of volume
• Not done much since OOPP to help with plugin issues
• Have not improved our ability to deploy blocklist
• Not made a blocklist tool for handling large problems
• Mobile can't persist state
• Socorro not good at adding new reports
• Socorro - lots of shifting priorities, creating churn
• More community involvement
• bad bucketing
• network issues
• not setup to scale easily
• crash investigation very labor intensive - need to automate more things
• classes of crashes that are getting little attention

Ideas

• Mailing list for crashkill
• Sorted bug list by Socorro rank
• Notification to users crashing a lot
• Crash rating system