The current system of a wiki page per incident on wikitech is a good start and captures most of the needed information but makes reporting and analysis very difficult.
Let's work through some options of what to use instead of or in addition to the wiki pages.
| Status | Subtype | Assigned | Task | ||
|---|---|---|---|---|---|
| Resolved | • mmodell | T208632 Outline options for managing incident reports creation, follow-ups, and analysis | |||
| Resolved | zeljkofilipin | T206622 Develop set of metrics to assess incident reports/post mortems |
One option is crafting some scripts to scrape the information from the wiki pages and putting that info into eg a Google spreadsheet for analysis. Zeljko already has some of this written for his current analysis (T199133).
Another option is using a self-hosted (eg: on Cloud VPS) (Semantic) MediaWiki install (see Cindy's install scripts at https://github.com/cicalese/debian_wikifarm ) where we could have all incident reports created using some form and then get analysis for cheap(er).
A crazy idea was to use Wikidata entities, but wikitech doesn't have Wikibase Client.
Would it even be possible to store the data in wikidata? I'd assume it'd be out of scope and we'd need specialized properties that wikidata doesn't have.
I've created a Ruby project zeljkofilipin/incident-documentation that extracts some information from the incident reports. Please see readme for the information it currently collects and issues for planned improvements.
The same tool (or a similar one) could be used to create incident reports for train deployments. Each train has a task, they have subtasks with tags and gerrit patches...
Options outlined in this task commentary.
For now we will continue with the current system (including @zeljkofilipin 's analysis scripts).