It would be nice if we auto-packaged up backtraces so they could be easily manually (Or automatically if an organization allows) be sent to support lines for debugging. Clang, GCC both have this sort of behavior built into their drivers. We could also borrow from web browsers potentially for this sort of thing, such as Mozilla's Breakpad (backtrace grabber) [1], and Socorro (stats viewing) [2].
[1] https://wiki.mozilla.org/Breakpad [2] https://wiki.mozilla.org/Socorro On Mon, Feb 9, 2015 at 11:55 AM, Niklas Nielsen <[email protected]> wrote: > Hi everyone, > > Has anyone worked on crash report tooling for Mesos clusters? > As a part of our testing (both internally and for testing of public RC's) > been looking at tools we can run to 1) monitor running mesos processes > (unusual behavior in use of file descriptors, and cpu load etc) 2) To grab > information in a post-mortem manner (crawl endpoints, find the command line > arguments processes were started with, machine stats and so on). > > Even if different organizations use different tools to do this, it could be > awesome to join forces and find a common format. > > Do you guys have any thoughts or ideas? We could host such a tool as a part > of the mesos distribution or just host it in github.com/mesos > > Cheers, > Niklas >
