Would metrics on kernel warnings be useful?
stefan.bader at canonical.com
Fri Nov 9 13:03:59 UTC 2012
On 08.11.2012 13:53, Evan Dandrea wrote:
> I had a chat with Stefan at the UDS after party and he suggested that
> kernel warnings might be a useful data point to capture in the error
> tracker (http://wiki.ubuntu.com/ErrorTracker).
WARNING actually was not the primary thing. More the info about kernel crashes
and how much of it. :) But WARNINGs at least used to be collected as well and
maybe we sort of mixed up talking about changes from kexec-tools to kdump-tools
and what kerneloops was done.
> I would be willing to implement this once we have the ability to send
> non-critical information with the next regular crash report. This way,
> users do not get a dialog each time this happens.
> So, would this be useful? :)
> Would the "WARNING:" line alone be sufficient, or would you also want a
> list of modules, log files, etc?
> Is the "WARNING:" line a sufficiently unique string to bucket instances of
> the same warning together?
I suppose the kerneloops package would be the place to look (and potentially we
should retire/kill it after having a new place). There are actually a few
messages (as documented in the changelog) that have WARNING but are not really
bad enough that one should gather them.
And no, there is various things that say WARNING. Which probably means one would
need to sort uniquely by the whole line.
-------------- next part --------------
A non-text attachment was scrubbed...
Size: 897 bytes
Desc: OpenPGP digital signature
More information about the kernel-team