High CPU load / unresponsive OH since upgrade (even with 5.0.2)

BananaBender · November 2, 2025, 5:46pm

Since upgrade to OH 5, my installation encounters performance issues, so that I’ve to restart OH latestly every second day. First, I thought, that this is the same like mentioned in

High CPU Usage since upgrade to 5.0.1 resp.

github.com/openhab/openhab-core

Persistence extensions: fix potential infinite loop

main ← mherwege:persistence_loop

opened 04:17PM - 10 Sep 25 UTC

mherwege

+144 -59

If an item has no values persisted, the `lastUpdate`, `lastChange` and `previous…State` persistence extensions could get into an infinite loop. This could also happen when the query to the persistence service fails when called from these extension methods, e.g. because the database is offline. Resolves https://github.com/openhab/openhab-addons/issues/19313 See https://community.openhab.org/t/high-cpu-usage-since-upgrade-to-5-0-1/165989/20?u=mherwege See https://community.openhab.org/t/high-cpu-usage-when-calling-lastupdate-on-offline-influxdb-database/166198 Also see https://community.openhab.org/t/averagesince-calculation-wrong-if-no-measurements-in-timerange-available/166177/12. This PR adds some tests that look for the described behaviour in the community post. The tests did not reveal the issue so far. If accepted, this should be backported. @Nadahar @jimtng FYI

but the fix to 5.0.2 did not solve that behaviour.

After 1-2 days runtime, my system (RPi 4B / OH 5.0.2) is unresponsive. Seems to be a problem with data persistence / rrd4j - ttop shows, that OH-MapDbPersistenceService remains in waiting resp. blocking CPU ressource.

Are there any other facing a similar problem?

Thx!!

Nadahar · November 2, 2025, 5:54pm

We need a thread dump to have an idea of what’s going on (the thread dump must be made when the problem is there, what OH is “stuck” or very slow)

openHAB 5.1.0Mx: UPNP Issues

After some testing on my openHABian, I found this approach quite easy:
sudo ps -ax | grep bin/java
It will list a lot of stuff, what you’re interested in is the PID of the Java process running OH, which should be the very first thing it prints (in my case it was 726).

Then, once you have your PID, run (insert your PID number)
sudo jstack -l 726 > threaddump.txt
The thread dump is then in the file threaddump.txt in the folder you’re in when you run the command.

BananaBender · November 2, 2025, 8:12pm

Thx. I’ll check it tomorrow resp. the day after, when the problem is back again after today’s restart.

mstormi · November 2, 2025, 9:52pm

You failed to provide details of your setup.
If you run persistence off your SD card, you might be affected by wearout or lame card I/O.
Do you use ZRAM?

Have you ever ramped up persistence logging?
You might be seeing messages pointing to the problem.
I for instance used to have corrupted rrd4j files that OH failed to write to.
Deleting / overwriting them solved my problem.