[Wiki] HTTP Features and Endpoints

Ken_Fraleigh · September 2, 2020, 5:41pm

I expect it was running a test while stats were on. I would hope 5 seconds would be the longest thing running (I’m assuming synchronously). It didn’t show up for me, but wasn’t running a test at the time. I only ran it for a minute.

Tony · September 2, 2020, 6:14pm

Yes, the picture may certainly look different over a longer observation period. Hence the need to go about this in some kind of structured fashion. It's great to have the ability to see what levers can be pulled and compare the outcomes.

lewis.heidrick · September 2, 2020, 6:25pm

I'm sure someone will come up with some sort way to prettify it into an app.

Tony · September 2, 2020, 6:32pm

Relative to my post above, these were my high flyers:

App stats total run time: 1094770
Apps:
runcount 78 total runtime 202420 average run time 2595.1282051282 << Hub Watchdog
runcount 202 total runtime 10063 average run time 49.8168316832 << HubConnect (ST Client)

Devices:
runcount 78 total runtime 1565 average run time 20.0641025641 << Environment sensor
runcount 60 total runtime 2256 average run time 37.6 << Zooz Powerstrip

Also: Memory GREEN 481476

Looking forward to seeing what I can tune here, now that I easily gauge my progress.

lewis.heidrick · September 2, 2020, 6:36pm

I'd turn down the frequency rate on Hub Watchdog until you start having problems. Mine runs every 3 hours.

I wonder why your power strip is so much higher avg run time than mine.

device id 1218 runcount 97 total runtime 611 average run time 6.2989690722
device id 1219 runcount 97 total runtime 558 average run time 5.7525773196
device id 1220 runcount 97 total runtime 530 average run time 5.4639175258
device id 1221 runcount 97 total runtime 483 average run time 4.9793814433
device id 1222 runcount 97 total runtime 498 average run time 5.1340206186
device id 1223 runcount 97 total runtime 461 average run time 4.7525773196
device id 1224 runcount 97 total runtime 443 average run time 4.5670103093
device id 1225 runcount 97 total runtime 459 average run time 4.7319587629

lewis.heidrick · September 2, 2020, 6:38pm

Fast is good for making pretty graphs but if it impacts your hub you're kinda shooting yourself in the foot.

Tony · September 2, 2020, 6:41pm

Lol, for sure. That jumped out at me like a sore thumb... but probably with a granularity of minutes rather than hours. Actually I'm going to disable it entirely for a while to see what kind of uptime the latest firmware can provide. I only started using it when the need for some kind of objective measurement of the progressive slowdown phenomenon became critical, last Fall.

lewis.heidrick · September 2, 2020, 6:42pm

Yea I only ramp it up to minute increments when I'm troubleshooting something. I keep it at 3 hours so I can still watch if it starts trending up. .142 has been stable for me. Haven't had to reboot and not currently trending up in delay.

Evilborg · September 2, 2020, 7:05pm

Correct

Ken_Fraleigh · September 2, 2020, 7:30pm

How many devices? My memory is half that on one hub and a quarter of that on the other. 129928 and 209976. Incidentally, the one with less free memory is faster so far.

Tony · September 2, 2020, 7:54pm

Roughly 75 physical devices (50 Zigbee, rest Z-Wave), 109 shadowed to ST via HubConnect; WebCoRE (107 pistons); 3 LAN devices, 59 virtual, 45 Simple Automation rules and a half dozen each of Motion Lighting and Button Controller instances. Plus a couple of groovy apps (Laundry monitor port from ST, HubWatchdog). RM2.5 and RM4 installed, but disabled. All running like a top, for now.

Is your memory status green, as well?
Edit: just checked and memory now 498424, up a tad. GC is doing its job...

kevin · September 2, 2020, 8:01pm

Bear in mind guys that an app is typically dormant until an event of significance to it is thrown. This is why big apps are not necessarily resource hungry.

However if your app subscribes to device events it will run whenever a relevant device changes state. Typically you select these devices within the app. So these (runcounts) look potentially more culpable than they are, but to function they must do this.

More important is how efficiently the app deals with those events before exiting again, so runtime is interesting.

MQTT publishing status updates to a broker or monitoring a broker for updates may run frequently but hopefully handle such events efficiently.

Tony · September 2, 2020, 8:04pm

That's why the run counts and total run time are interesting (and necessary) to see the big picture.

syepes · September 2, 2020, 8:10pm

I see you guys are hungry for information, so I would like to contribute with the the list of REST endpoints I have been able to find over the past years. Some response with GET or POST and some are no longer working with the last versions but but I am sure some of you will dig deeper.

Filtered out the dangerous ones

lewis.heidrick · September 2, 2020, 8:11pm

I get it. I work in process automation in a refinery on 10-20 year old hardware.

lewis.heidrick · September 2, 2020, 8:12pm

Capturing this page before someone deletes the post.

Evilborg · September 2, 2020, 8:30pm

Someone has seen the FW code...

neonturbo · September 2, 2020, 8:42pm

I will be the nay-sayer here, and warn people that they shouldn't probably mess around with this unless you know what it does, or are instructed by support to do so.

I am just picturing the "my hub is bricked" threads that are going to be littering the forums.

syepes · September 2, 2020, 8:45pm

Exactly, don't test any of this in your main / production Hub... I can assure you your wife won't be happy

Evilborg · September 2, 2020, 8:45pm

I'm going to step out on a branch and say this probably wouldn't have happened if the HE people had been more forthcoming especially when they aren't giving proper attribution to the open source code/projects they are using which is a license violation. The more they resist the more people will push back.... The thirst for knowledge is strong.

But I do agree..... playing with some of these url's has the potential to brick a hub so be very careful.