openHAB 4.1.0 Milestone 5 - Docker - Slow + 100% CPU load

[SHU-red]

openHAB 4.1.0 Milestone 5 - Docker - Slow + 100% CPU load

[SHU-red]

Going back to the openhab:latest image containing openHAB 4.0.4 solves the issue

[kaikreuzer]

The info log entries are coming from the network binding. Could you check if the behavior disappears if you disable the network binding things temporarily (or deinstall the network binding completely)?

[SHU-red]

I did the following

1. openHAB 4.0.4 deactivate all network things via GUI
2. Re-Deployd openHAB M5
   CPU load much lower

Seems to be the root cause

Maybe off topic:
Now the frontail log was completely bloated with many of the following errors
(Not sure if this is a migration-thing?)

Script execution of rule with UID 'counters-3' failed: An error occurred during the script execution: Cannot invoke "org.openhab.core.model.rule.scoping.RulesClassCache.get(Object)" because "cache" is null in counters

EDIT:
Additionally

• openHAB did not seem to run stable
• after stopping and starting the portainer-stack it wont come up again with milestone

[kaikreuzer]

Good to hear that the CPU load issue is located and that it is not a general issue in core. I'm not sure what might cause this in the network binding, though. We would probably need to have a closer look at what kind of things you have exactly configured there. I'd suggest to raise an issue for the network binding in openhab-addons.

Wrt the other comments: All of these would need more analysis and details before anything can be done about them.

[SHU-red]

Sorry @kaikreuzer,
my morning was kind of hectic and i think i gave you wrong informations

Even after Inactive Network devices and deinstalling the Network Add-on i get

Either openHAB not booting at all --> Low CPU
Or booting and causing high cpu utilization

If i can do anything regarding debugging i will try to do so

[miloit]

I can also confirm same behavior on my raspberry pi 4

[kaikreuzer]

@miloit Also on docker?

@SHU-red Could you try to disable the three "suggestion finders" under Settings->System Settings->Add-on Management->Show advanced?

@wborn Do you notice any general issue with M5 in docker?

@mherwege & @andrewfg Do you think the suggestion finders could have any side effects here when using docker?

[andrewfg]

Do you think the suggestion finders could have any side effects here when using docker?

Let me answer in several parts..

• It's unlikely that the mdns scanner would have any impact at all since mdns is anyway already running.
• I personally don't see any impact potential from the upnp scanner. But insofar as others already complained about upnp, those kind of scenarios would also have a similar impact (or not) on docker. I.e no change expected from prior.
• I can't comment on the process scanner, but AFAIK it is a one shot scan. Rather than repetitive. ??
• I can't comment on the IP pinger scanner either. But it is using UDP so unlikely to affect HTTP. Or?

[miloit]

I am on obenhabian with raspberry 4..

[mherwege]

• I can't comment on the process scanner, but AFAIK it is a one shot scan. Rather than repetitive. ??

It scans when suggestions are retrieved from the REST api. That would be when opening the UI setup wizard or add-on store. So I cannot see impact at startup.

The IP scan finder scans only once at startup.

The process finder cannot be disabled. The other 3 can through an advanced option for add-ons. Because of this, they are only installed (feature installed) in the suggestion finder service activation. Could installing these features at that time have this impact? This is similar to installing add-ons. Disabling these finders should allow to see this. Also, switching on trace logging for org.openhab.core.config.discovery.addon could show the sequence and timing of these feature installations.
I don’t have experience with docker, so cannot test myself. I am not convinced that’s the reason, but it could be worth checking out.

[holgerfriedrich]

I cannot reproduce this behavior on 4.1.0.M5 (without Docker on RPI4 openhabian, and with Docker on Windows).
I created the container a few times and restarted as well. I experience normal CPU usage.

As Kai said, it seems the message

CIDR prefix is smaller than /24 on interface with address .../16, truncating to /24, some addresses might be lost

is from the network binding. If I uninstall the binding, the message is gone.

It seems we need you help to debug.

[J-N-K]

Indeed the IP scan in the network binding is problematic. I remember that I introduced the code that prints this warning and IIRC the issue is that the underlying code generates a list of all possible IP addresses before the scan starts. A /16 subnet has 65534 hosts, so a lot of memory is allocated and eventually lead to an OOM exception on low-memory systems. It might well be that the code also blocks a thread while scanning and that is an issue even for /24 subnets during startup.

[SHU-red]

There is something that comes to my mind which could be connected to this problem:

• I recently got the error on docker that i reached the limit of 30 Docker Networks
• In this case there is a fix to extend the "address-range" used by docker proposed here: https://stackoverflow.com/questions/41609998/how-to-increase-maximum-docker-network-on-one-server
• I added the *daemon.json file to my coreOS which made it possible for me to create more than 30 containers

Could this be the reason why my system causes problems (by scanning a huge huge amount of addresses), compared to to other ones?

Im not sure what will happen if i make changes to this file wile running many containers but i could maybe take the time to stop all of my containers, re-configure the address-range and re-test M5

But this is just a guess

[wborn]

Do you notice any general issue with M5 in docker?

Works fine for me so far.

I think it is a known issue that when using the network binding with Docker it scans too many interfaces resulting it a lot of load, see also:

• https://community.openhab.org/t/openhab-4-0-1-and-network-pingdevice-on-docker-install-results-in-massive-threads/148658
• openhab/openhab-addons#15437

[SHU-red]

I think it is a known issue that when using the network binding with Docker it scans too many interfaces

Sounds like could be the thing! Thanks

[kaikreuzer]

So can we close this issue then, @SHU-red?

[SHU-red]

Im not absolutely sure 😅
It looks like i will face problems as soon as i switch to a release of M5 or newer ...

But maybe you want to propose that we close this issue and track the debugging in one of the other already existing issues --> Makes sense

Are the other known things also related to especially M5?
I think in my case M4 was running perfect but especially M5 is bringing these problems --> Just wanted to make sure that the topics are really about the same root cause

[kaikreuzer]

Yes, I would have thought that it can be further tracked under openhab/openhab-addons#15437.

But why you never had the problem and it only turns up now with M5 is indeed a mystery...

[SHU-red]

Ok,
whatever is senseful in your mind is ok for me

To morrow I will post a comment on the other topic
If i can help tracking the the problem just let me know

Thank you for taking the time

[SHU-red]

Closed due to assumption that root cause is the same as in
openhab/openhab-addons#15437