Elevated and Severe Hub Loads with v2.3.4

Ping @gopher.ny with a DM and provide your Hub ID and he'll pull the engineering logs.

There is no fix atm, that I'm aware of, - however, in the meantime, here is my RM 5.1 rule that automatically reboots my hub after 10 Severe Load Alerts. This stops your hub from becoming completely unresponsive IME.

https://www.dropbox.com/s/myzvnyjrfpb1e3q/_Hub%20-%20Severe%20Load%20Mgmt.json?dl=0

I did.

Thanks. I already have a RM rule that notifies me if there's an alert. I plan on adding your code to mine. (Saw it earlier but haven't gotten around to it yet.)

It's not happening for me as often as it is for others, so for now I'm keeping an eye on it. But I've had my hub — it's a C-5 — for a long time and never had this happen. I guess my next straw to grasp at would be to remove the HomeKit integration. That's when it seemed to start. But that also might just be coincidence.

1 Like

I removed HomeKit from my secondary hub and it made no difference - it's more likely to be a platform issue than a HK issue IMO.

PS, I've just been offered a beta to test, so I'll report back on how that goes.

1 Like

I had this for the first time this morning. It almost looks like the elevated CPU is directly related to the lower free mem. It has not done it before but I was rebooting alot due to updates. This was the longest uptime I have had in a while.

That matches my observations too. Below 220 KB (or is it MB?) is the danger zone IME.

I had it happen overnight on both hubs and free memory was over 320 on one and 360 on the other.

I just had the elevated load on my C5 tonight. I haven't seen it before and was on .134. Apps and devices were fine. It has to be an internal process. I rolled back to .130 and will see what happens.

Victor let me know that one of the fixes coming, is the switch to TCP, and removal of UDP, for hub mesh.

I’ve noticed some improvement from making that switch on .134

1 Like

It looks like the team finally got to the bottom of the issue. I’m beta testing .136 and it’s looking really, really good!

Compared to .134

2 Likes

I think you want to compare uptime apples to apples to really judge memory. The JVM does a lot of stuff 'over time'

3 Likes

I updated to .136 earlier. It's down to 348000KB after just 6 hours. I wasn't going to bother updating from .134 but that had dropped below 200000 after a few days.

I will say that seeing ongoing issues for some with high cpu/low memory is making me over analyse it. However I remember going on holiday for 8 days in September and my hub was still at 350000 on my return after 11 days uptime.

1 Like

I wasn’t intending to compare memory, I was comparing CPU loads. My primary hub is seeing greatly reduced loads, even after only an hour or so of runtime.

Memory does still drop quite quickly initially, but it seems to have stabilised on both hubs.

136 must be beta

2 Likes

It is.

Oops - “the first rule of beta is, you do not talk about beta”

12 Likes

Posted this in another thread.

But ever since the last two firmware upgrades having the same elevated and severe hub load issue after firmware updates .133 and .134.

I have narrowed it down to occurring daily to the hour after a reboot. Also seems to coincide with a drop in memory to a threshold of around 250 mb.

All device and app logs have remained consistent at a pretty low % before and after this issue and firmware updates.

Are you setting the hub to reboot for any particular reason?

The reboot is manual. Because of the constant notification of sever hub load "that may result in performance issues or radios being shut off".

If I ignore the issue, at some point I fear that I will have performance issues that will have cause to reboot the system anyway.

I am on latest firmware, and only have 4 z-wave smoke/co detectors on the hub, all have been integrated into HomeKit. I am also getting Hub Load Severe notification. When I look the log, other than the detector status check and heartbeat logs, there is nothing else that indicates any errors, was this a normal situation or there is something else that I have missed? So far, I have not experienced any malfunction from HomeKit point of view. Any advise and insight of this will be greatly appreciated.