#No no errors or warnings on the console
obsidian blaze
I’ve got one dashboard that’s just auto-entities generating stock entities cards. And I can tap a beat with my foot as each card appears.
Like 0.5s between each card appearing.
sour dagger
hmmm...
a) try going to "developer tools" > "event" tab
b) tap asterix * in "Event to subscribe to" and "Start listening" button
c) you should see events appearing in the log
(this log only holds 50 last events... but it's interesting how fast new entries arrive)
obsidian blaze
so just did a quick test. Subscribing to * events and watching it for a minute resulted in about 600 events. But I have no idea if thats a lot or not.
sour dagger
quite a lot (10x then my system with 120 devices) but probably not enough to tank performance completely... that's still ~ 1 state / 100ms on average
it most hardware should handle that
Not sure how large your sys. is but If there are particular entities repeating in majority of those events then disabling them will give some info on perfomance impact
other then that I would recommend playing with the performance tab in browser dev tools
it's the only sure way
obsidian blaze
So I haven’t had a chance to sit down with the browser dev tools yet, but just following your message…. Does every event that is seen in the HA dev tools get sent through to the browser as well??
sour dagger
Usually "state_change" events - but those make over 90% of traffic usually.
Some other event types only if they are listened to specifically.
The traffic can be monitored in the network tab
(filter by "websocket" and check response tab - this connection is being kept open as long as tab is loaded)
Still this does not translate 1:1 on performance impact.
e.g. most cards are conditionally watching particular changes (usually related to entities being displayed) so they are "resistant".
Some, like Auto Entities react to any change so they are affected.
600 state changes per minute doesn't sound that bad. I emulated additional events on my system, and while small cpu increase was observed my dashboard (and whore UI) remained as snappy as always.
While performance it's very relative to your hardware and browser engine I don't expect this to be the sole reason why it tanked this bad.
Anyway... is the performance drop noticeable only when dashboard is loaded, not on any other view?
(measured on freshly reloaded browser tab)
obsidian blaze
Thanks for the additional info, its very helpful.