Hacker News Reader

Comment by emj | Hacker News Reader

Matomo is nice on low traffic, but when we have a sustained rate of 5-25 logins per second and above things become real slow. Using regexps is really bad when you start having problems, but they are fine on low traffic sites.

So If like it but it is a headache on high traffic sites. If anyone have an easy solution I would gladly accept it.

View on HN

I have worked with two clients. Both north of 8 million visits a month. Both on matomo. Both self hosted.

If you architect the underlying infra right it still works like a charm. But I admit people need to know what they are doing. I was quite impressed with both infra teams.

But as always, if you do not want tu use auto scaling US cloud based services, you need to enasure you have the right scaling and the necessary technical expertise at hand.

I had no problems either, until we hit peaks. We hit our problems at about 7 million unique logins per month, we do not track visits in the same way. I am not that invested in Matomo and it just costs time for me.

I am not sure how you scale Matomo we could not vertically scale anymore, we never did MySQL clusters because it just was not cost efficient for internal reasons.

Managed a fairly large matomo site in the past. Using queue plugin (https://plugins.matomo.org/QueuedTracking) with Redis Cluster really improves the situation. We actually built a custom plugin with Nginx + Lua to avoid PHP altogether for the tracking part. Scaling ingestion then wasn't the problem, draining the queue was

The tracking was not the issue the problem was report generation with segments. Every segment makes you regenerate all the reports. Tracking part is a problem because you need to split the tracking and report part of you want to have something robust.

Use clickhouse

Reply on news.ycombinator.com