lpnet reports fail when things are changed

What about giving it a list of teams instead of prefixes? I think there's already code to track prefixes for each team, and it would avoid the need to specify the prefixes in more than one place.

On Fri, Mar 25, 2011 at 7:24 AM, Selene Scriven
<email address hidden> wrote:
> What about giving it a list of teams instead of prefixes?  I think
> there's already code to track prefixes for each team, and it would avoid
> the need to specify the prefixes in more than one place.

At least for launchpad, we no longer split things out by team : we've
removed the internal silos and driven the frequency down low enough
that any oops is relevant to anyone doing maintenance (vs feature
development). If doing it by team is cleaner we could indeed do that.

I'd like to remove the analyse script. It's not being used anymore. Currently we have the following reports:

 * production-YYYY-MM-DD
 * staging-YYYY-MM-DD
 * qastaging-YYYY-MM-DD
 * soyuz-YYYY-MM-DD
 * production-last-hour

The initial 4 are sent daily to the list and contain the oopses from the previous day. The last one is the data source for the graph: https://lpstats.canonical.com/graphs/OopsLpnetHourly .

Those reports and their respective prefixes are configured here: https://lp-oops.canonical.com/admin/oops/report/ and generated by the scripts: report.py and graph_report.py

I'd like to remove the code to create reports by team as well, since that never really worked well and it adds more confusion than helps.

On Sat, Apr 16, 2011 at 12:54 AM, Diogo Matsubara
<email address hidden> wrote:
> I'd like to remove the analyse script. It's not being used anymore.
> Currently we have the following reports:
>
>  * production-YYYY-MM-DD
>  * staging-YYYY-MM-DD
>  * qastaging-YYYY-MM-DD
>  * soyuz-YYYY-MM-DD

I believe we can drop soyuz now; I need to double check with bigjools
but he was amenable to the idea once our oops counts were low enough
to sanely browse a single report.

>  * production-last-hour
>
> The initial 4 are sent daily to the list and contain the oopses from the
> previous day. The last one is the data source for the graph:
> https://lpstats.canonical.com/graphs/OopsLpnetHourly .
>
> Those reports  and their respective prefixes are configured here: https
> ://lp-oops.canonical.com/admin/oops/report/ and generated by the
> scripts: report.py and graph_report.py
>
> I'd like to remove the code to create reports by team as well, since
> that never really worked well and it adds more confusion than helps.

+1

FWIW, ISD is still using the analyse_error_reports script. It is part of a daily cron script which sends that day's errors to our mailing list. The on-demand web reports are nice too, but I'd like to keep some way to dump the summary to stdout at a command line.

One way would be to use an API client to drop the summary to stdout. I'd like to have oops-tools integrated into the LP web UI, though thats a little down the track.