On Fri, Aug 14, 2009 at 4:27 PM, Tim Steele<[email protected]> wrote: > I like the idea of closing the tree automatically more often. Especially > for rookies like me, reacting early and consistently is a way lower stress > way to be sheriff than permitting redness. > Two days drinking the pkasting kool-aid on the front line and I know that I > will run a tight ship the next time duty calls :) > On Fri, Aug 14, 2009 at 12:29 PM, Amanda Walker <[email protected]> wrote: >> >> While bot health is separate from tree health, I'm not sure we really >> want to tolerate redness because a bot is sick--it'll still keep us >> from noticing regressions or bustage in a timely fashion. However, a >> way to visually distinguish between "build failure" and "bot failure" >> would be nice. "out of disk space" should be easy to detect--is there >> any reliable way to detect purify running out of RAM, or other bot >> failure modes we know about? > > Last week there were these weird timeout issues with 'compile sandbox' on > some of the bots, like Modules Vista; not sure what was happening there. > But it was one of the few things we let slip through, so I remember it. >
Buildbot glossary: Builder = column in the waterfall Slave = a machine connected to a builder. So s/bot/slave/ By definition, slave failures are unpredictable. If they are predictable, it must be fixed anyway. :) I'll update http://dev.chromium.org/developers/tree-sheriffs once the "ignore builder" functionality exists. M-A --~--~---------~--~----~------------~-------~--~----~ Chromium Developers mailing list: [email protected] View archives, change email options, or unsubscribe: http://groups.google.com/group/chromium-dev -~----------~----~----~----~------~----~------~--~---
