Actually, the normal "hits" data is sometimes unavailable, too. Here's the same date range (May 2022 - May 2023) for hit counts across all tools:

I'm assuming getting historical data for the missing hits and unique visitor counts won't be possible, so I can state in the new Toolviews frontend that some data before [some date] may be unavailable. Some sort of user-facing warning should be shown, I think.

In T338053#8899324, @MusikAnimal wrote:

I'm assuming getting historical data for the missing hits and unique visitor counts won't be possible, so I can state in the new Toolviews frontend that some data before [some date] may be unavailable. Some sort of user-facing warning should be shown, I think.

You are correct that we do not have any way to backfill missing data. The database is populated by logrotate triggers on the proxy servers. There are a number of reasons that data might not make it to the database, but the most likely one would be issues with ToolsDB itself at the time that a front proxy's nginx access log was rotated. We do not currently monitor or alert on failures in this pipeline.

Unique visitors numbers do seem to be very broken. That's a feature that @Andrew added to the system. My only real involvement with it was merging a patch on the API side of things. I'm not seeing any data at all in the daily_ip_views table and I think that if I understand the general logic of https://gerrit.wikimedia.org/r/plugins/gitiles/operations/puppet/+/refs/heads/production/modules/toolforge/files/toolviews.py that table should nearly always have some data in it.