Proxmox LCX, Debian 12 (VM on a Synology NAS) - openHAB 4.3 - also tried Milestone OH 5
Hello!
at the moment i am transferring my Servers away from Docker / Synology to Proxmox.
I am having mixed success but openHAB didn’t cause that much problems - until yesterday at around 0 am. The CPU load increased to + 50 % and the ram maxed out.
Also the KNX binding was not working anymore (failed to connect to the router via tunnel).
I rolled back to a backup and everything worked fine until 60
min ago - same behaviour - knx binding also offline. However when i tried Router instead of Bridge it worked but the CPU load and Ram usage is still very bad.
At the time I was at the Milestone release - but then i tried going back to the stable version 4.3.
I did some digging and found out that the heap size can cause issues so here’s my current values:
openhab> shell:info
Karaf
Karaf version 4.4.6
Karaf home /usr/share/openhab/runtime
Karaf base /var/lib/openhab
OSGi Framework org.eclipse.osgi-3.18.0.v20220516-2155
JVM
Java Virtual Machine OpenJDK 64-Bit Server VM version 21.0.6+7-Debian-1
Version 21.0.6
Vendor Debian
Pid 5360
Uptime 14 minutes
Process CPU time 10 minutes
Process CPU load 0.34
System CPU load 1.00
Open file descriptors 188
Max file descriptors 102,642
Total compile time 4 minutes
Threads
Live threads 143
Daemon threads 77
Peak 147
Total started 612
Memory
Current heap size 1,733,114 kbytes
Maximum heap size 2,035,712 kbytes
Committed heap size 1,994,752 kbytes
Pending objects 0
Garbage collector Name = 'G1 Young Generation', Collections = 261, Time = 31.020 seconds
Garbage collector Name = 'G1 Concurrent GC', Collections = 90, Time = 5.948 seconds
Garbage collector Name = 'G1 Old Generation', Collections = 0, Time = 0.000 seconds
Classes
Current classes loaded 21,716
Total classes loaded 21,893
Total classes unloaded 177
Operating system
Name Linux version 6.8.12-8-pve
Architecture amd64
Processors 2
Total physical memory 8,135,388 kbytes
Free physical memory 672,128 kbytes
Okay …
Still i cannot get it to work …
I experience 100 % CPU Load and 100 % RAM usage even after a brand new installation of the stable build with openhabian. I tried a debian 12 and ubuntu 24 lxc.
I rolled back to my synology install (docker) since also on docker (inside an lxc) i had the same issues.
I am feeling pretty bad not coming any closer to a solution after XX hours of trying.
By the way: That is the main reason why i am transferring away from synology since the installed reverse proxy locks port 80 so i cannot bind that to traefik for certificate renewal. and so i cannot use authelia. another reason is the lack of ipv6 support for docker on synology.
I guess it was the SLAAC setting in the Network-Settings (I enabled because the matter binding wasn’t working) but it seems to overload the ipv6 mdns service … Uptime is only 20 min now but this was enough before to drive the load and ram up to 100 % … thanks for the hint with the ttop --stats=tid,name,state,user_time,cpu_time,user_time_perc,cpu_time_perc --order=cpu_time --millis=300 → this showed clearly the problem with mdns.
Nice! Why does it only have one CPU? Might be a bit restrictive for an application like OH that has lots of threads. On the Synology it probably had more CPUs available?
yes. I just gave it one for testing (my productive environment of OH also runs on docker also on the synology) and since it went to 100 % every time i restricted it to one core so my system doesn’t freeze completely. I have 4 cores available and allow it 2 - which is enough normally. i had to set it to static. i tried dhcp but was not able to log into the console from the proxmox. i dont’t know if dhcp ipv6 is required for matter though … my router and dhcp server (unifi usg) does not support ipv6 at all i think.
Yeah IPV6 is mandatory. I just don’t know if it will work with a static ipv6 my router doesn’t support ipv6 - i’ll give it a try.
I guess so because i got it to work last week. but even if i don’t - better a stable OH than getting matter to work
Thanks for the hint
