Database crash + queued tracking = where's the data?


(Duncan) #1

Our Matomo database crashed at some point over the weekend. We use redis-backed queued tracking.

When we noticed the issue this morning, we noted 2GB worth of requests in redis. We restarted the database and it quickly ingested the queue. We then invalidated the reports, and checked back a few hours later, but it seems none of cached data is included. It still shows Sunday as 0 for traffic, and today is significantly short. This happened once other time in the past and the data never did end up getting recorded it seemed.

Perhaps I’m mistaken, but I though the queue tracking was designed for this sort of occurrence. Is there something I’m missing?


(Duncan) #2

Sorry to bump this, but the issue occurred again and I’m desperate to figure out if our redis-backed queue is working correctly.

(Matthieu Aubry) #3

Can you check that the data is correctly tracked, by using Visits Log on the crash day and check the data was ingested correctly?

if it is, then probably the issue is the invalidation. maybe it didn’t invalidate. What command did you use?