Like I said above, I had the same problem with Maker slowing my hub to a crawl. So it isn't just one person. Admittedly Maker in my case wasn't being used with Node Red, but I am not sure that makes any difference?
I suspect that others are also having issues too. I have seen many instances of people saying they use Maker and they are having general slowness issues or having to constantly reboot and so on. But they aren't connecting the slowdown with Maker for some reason.
Anyhow, I strongly suspect there is some bug, memory leak, or whatever in Maker. Maybe only when used a certain way, or only enough of a bug when combined with something specific (a built in OR community app/driver?) that it becomes apparent.
I use Maker API on 3 hubs, and haven't seen it. BUT I also use it more or less the same way on all 3 hubs. So if it only triggers in some uses cases, or with some certain devices, I might not see it. (3 of basically the same is really only "1" from a testing scenario).
Even with the slow decline in memory (aging sucks haha) my and my client's systems have been very responsive with Maker and NR - to be fair there are no other apps etc so maybe that offsets a bit. I have always suspected it's something related to "groups/scenes" but that's purely anecdotal and likely wrong.
Just on that one device alone - it was doing probably 20 or so a minute (multiple attirbutes involved) so 30k a day !!!! I've probably managed to cut it down to a thousand.
Curious now your graph looks on 2.2.5? Mine looked a lot like yours before: new graphs look different, but I don’t think it will fix reboot requirement on my C4.
You can just disable it with the X thing. It may level out though. A lot of that is just caching and whatnot. Especially after a reboot and about a week after that.
I've noticed that I have quite a bit more free memory since 2.2.5 rolled out (always seems to be over 500 now and before was always between 200 & 300). I wonder what @gopher.ny (or other HE staff) has done to cause this.
I would love to take credit for lower memory consumption, but in reality we just fine tuned what counts as free memory on each generation of hubs. In any case, memory value only becomes relevant if the hub slows down/freezes/crashes.
I let one of my C4's go critical in the interest of science.
Currently running 2.2.4.158 so not quite the latest. It was updated 2020-12-08 so lasted just over 2 months. During that time it continued chugging away with no noticeable loss in performance other than the web UI which became ridiculously slow about a month in. Like 20 - 30 seconds to navigate between pages.
It finally tanked last night, was still responding to ping but unresponsive on the main web UI AND on port 8081, so only option was to pull the power. I guess both the JVM and whatever serves port 8081 both died in the end?
Both my C4 hubs are only used for ZigBee (51 devices on this particular hub) and the only Apps are Maker API and a few small custom ones of my own. I use mostly system device drivers and my own so not much on there. Everything is in Node-Red using the event socket for real time updates and Maker API for "control".
Interestingly the last dying log entries were from Maker API, something about "excessive run", unfortunately I don't recall the exact error and "past logs" doesn't go back far enough now. I suspect that was just a final result of slowdown as opposed to the cause though.
I'll probably update to 2.2.5 in the next few days, but in the meantime @gopher.ny if there's anything on there that you want to look at feel free.
Graph below, to me it seems that the free JVM memory is mostly stable throughout, it's the free OS memory that declines steadily. I would say that when the free OS memory goes below the total JVM memory was roughly when the web UI slowdown started and steadily became slower as the free OS memory falls.
Yeah this is why I shut down Zigbee and Z-Wave services on my C-4 and repurposed it to be a "cloud"/Network device hub only. No apps other than Maker/Echo Skill/Homebridge V2 & Lutron. Decline is still there but it is very mild so I may put in a monthly reboot. As previously noted by HE staff the C-4 is running 64 bit which may actually work against it in terms of memory resources given the hardware constraints.
