mozilla-central/taskcluster/docker/system-symbols-linux-scraper/README

Source code

Revision control

Copy as Markdown

Other Tools

Performing a (re)bootstrapping of symbols scraping process
==========================================================
Whenever for any reason the symbol scraping process might have been faulty long
enough, we can end up (currently) in a situation where the recorded status of
`SHA256SUMS.zip` on the TaskCluster index is inconsistent with what we really
processed.
This document aims at explaining what needs to be done and where to recover
from that state (this is based on the experience from bug 1893156).
First, you need to identify since how long the problem has been present. As of
now there is no really better tooling than processing manually the cron tasks
logs and see when it started to fail.
Once you have identified a date, the next step is to work on the bootstrapping
content. As visible in
the first source of truth is the gh-pages branch of the symbol-scrapers github
source of truth is evaluated ONLY if the TaskCluster index is NOT present. The
route is being computed from the running task's definition:
from which we ONLY consider the `latest` alias.
As of today the index is for example for debian:
index.gecko.v2.mozilla-central.latest.system-symbols.debian and thus one can
explore the content at
other means of browsing including pushdate allows to find e.g.,
from which we can get a link to the sums file:
Once you have identified WHEN the problem arose, you can take the above URL
(adapting with the correct date) and adapting to the various distributions.
Make sure you have an uptodate git clone of the mozilla/symbol-scrapers
repository, checkout a new branch out of the gh-pages tree, and you can proceed
to the data extraction following (example with a different date):
for distro in alpine archlinux debian fedora firefox-flatpak firefox-snap gnome-sdk-snap mint opensuse ubuntu; do
done;
mv archlinux/SHA256SUMS.zip arch/
Please note that there's a slight difference in naming, archlinux vs arch.
Please note other distros might have been added since so you need to adapt.
Send a pull request once this is OK, make it reviewed or merge it.
As of now, the content of the boostrapping process is in the state we want, but
if you run a symbols scraping task, it will still pull data from the
TaskCluster index. This time, you need to use the index that refers to latest
and NOT the pushdate or another one, so the index used in the example SHOULD be
good. You just have to run deleteTask (with the appropriate credentials if you
have them, or ask releng for help in #firefox-ci):
for distro in alpine archlinux debian fedora firefox-flatpak firefox-snap gnome-sdk-snap mint opensuse ubuntu; do
taskcluster api index deleteTask gecko.v2.mozilla-central.latest.system-symbols.$distro
done;
From there, HTTP queries to (for the debian example)
would return 404, which will make the symbol scraping tasks search its data on
GitHub.