Page MenuHomePhabricator

As a stakeholder, I want to know how often images are failing to get sent to PhotoDNA
Closed, ResolvedPublic5 Estimated Story Points

Description

Building off https://phabricator.wikimedia.org/T287511, we need to run the MediaModeration script and count instances of the two debug messages:

And compare the two to find out how high the error rate.

First, we need to confirm the new debug code is on mwmaint by 2022-05-19, and if not, find out why.

Findings: Based on a grep of the logs, there were 574 images successfully sent to PhotoDNA out of 974156. This is a rate of .05% which seems pretty low.

Event Timeline

@jsn.sherman suggests using number of files rather than existing error messages.

We could maybe use Ellen's output

Or this message: Creating moderation request for file

ARamirez_WMF set the point value for this task to 5.May 16 2022, 3:33 PM

I think there are two fairly straightforward ways to get these numbers:

  • essex suggested that we sed/grep the logfile, which would be a really easy way to get started with no code change
  • if we want to have these values on an ongoing basis, we might be able to just accumulate the two counts in memory during a run and report them at the end, printing to console, and/or logging to file. so that we have the success rate for each run. This might be more useful if we want to have ongoing visibility and potentially raise some alerts for future automated runs

I was thinking the first approach. Creating the statistic on an ongoing basis is out of scope for this task, as we need to get these numbers sooner to know how to proceed.

I grepped the logs today and got 574 successes out of 971476 files sent.

Success message: "should be 3000"

Files sent message: "Creating moderation request for file"

mepps updated the task description. (Show Details)