Event limits not respected

@gopher.ny I just noticed both my C4's running don't appear to be respecting the event limits (not sure about states as it's difficult to check those).

Event / State limit is set to 11 / 11 respectively.

A possible bug?

C4 runtime statistics:

One device:

One device events page showing many humidity events .... way more than 11 (there' are a total of 567 entries which matches the total on the runtime statistics page)

The other C4 is the same behaviour so I've not shown screenshots for that, however my C7 running seems much closer to what I'd expect:

Please try running /hub/cleanupDatabase endpoint directly.
If that doesn't help, try rebooting and running it again.
If that still doesn't help, PM me with your hub id and I'll check it out.

I noticed this morning that the numbers on both C4 look much more normal so it looks like a cleanup may have happened over night anyway.

I thought the cleanup was supposed to be every hour though?

It is supposed to be every hour, but on that C4 it happens at 2:15am UTC only. Strange... I'll check it out. Is the hub slowing down noticeably in the meantime?

I'm seeing the same thing. The settings seems to be ignored.

I received a few warnings a while ago about DB size so I changed everything to 10/10 but I still get counts well beyond that. I didn't worry about it because I stopped seeing warnings about DB size.

10/10 causes an immediate cleanup after every event/state change, which turned out to do more harm than good on busier hubs. I'm recommending 11/11 instead these days. Although if it has no obvious side effects, no need to change it. The immediate cleanup logic is going away in 2.2.8.

1 Like

Thanks. I don't see any issues so I'll leave it for now since that logic is being removed. If I notice anything that might be related I'll try changing to 11/11 first.

It does seem to be overnight only on both those C4's, not sure when it started but the last restart on both was 12 days ago when I installed the latest release. I usually check runtime statistics first thing in the morning, so at that point the numbers probably looked reasonable. I think I only just noticed it this week as I'm on holiday so happened to check later in the evening and the red font stood out!

I've not noticed any slow downs, but none of my hubs are particularly loaded so typically I get 30 - 45 days or so now before memory becomes an issue and slow downs happen.