Memory usage of OH4 on RPi 4 model B within Docker

Setup, Configuration and Use Beginners
1
  • Platform information:
    • Hardware: Raspberry Pi 4 model B (4 GB RAM)
    • OS: Raspberry Pi OS / x64 / 12 (bookworm)
    • Java Runtime Environment: OpenJDK 64-Bit Server VM version 17.0.9+9-Debian-1deb11u1
    • openHAB version: 4.1.1 (running in Docker)
    • ConBee 2 USB stick
    • Output of openHAB’s console shell:info:
Karaf
  Karaf version               4.4.4
  Karaf home                  /openhab/runtime
  Karaf base                  /openhab/userdata
  OSGi Framework              org.eclipse.osgi-3.18.0.v20220516-2155

JVM
  Java Virtual Machine        OpenJDK 64-Bit Server VM version 17.0.9+9-Debian-1deb11u1
  Version                     17.0.9
  Vendor                      Debian
  Pid                         27
  Uptime                      5 days 23 hours
  Process CPU time            1 hour 24 minutes
  Process CPU load            0.00
  System CPU load             0.04
  Open file descriptors       230
  Max file descriptors        1,048,576
  Total compile time          14 minutes
Threads
  Live threads                193
  Daemon threads              112
  Peak                        204
  Total started               256941
Memory
  Current heap size           306,941 kbytes
  Maximum heap size           972,800 kbytes
  Committed heap size         481,280 kbytes
  Pending objects             0
  Garbage collector           Name = 'G1 Young Generation', Collections = 465, Time = 12.390 seconds
  Garbage collector           Name = 'G1 Old Generation', Collections = 0, Time = 0.000 seconds
Classes
  Current classes loaded      34,110
  Total classes loaded        43,396
  Total classes unloaded      9,286
Operating system
  Name                        Linux version 6.1.21-v8+
  Architecture                aarch64
  Processors                  4
  Total physical memory       3,885,396 kbytes
  Free physical memory        41,056 kbytes

Dear community,

I recently noticed that my Raspi reached its memory (i. e. RAM) limits and I wanted to know which apps consume the most.
I have several smaller apps that I developed myself and therefore I’m aware of their memory implications. However, openHAB’s memory footprint surprises me a bit - but of course, since I’m not familiar with the codebase, this is not a complaint :wink:

Here’s a htop screenshot of my system:

image
image1741×364 170 KB

As you can see, openHAB takes 1 GB. What I also noticed is that when changing the htop view so that threads are displayed rather than processes, there are approx. 100 threads related to openHAB. This is probably due to the following setting that I had to set within /opt/openhab/conf/services/runtime.cfg to get rid of openHAB’s startup error Error while connecting: Insufficient configured threads: required=11 < max=10 for QueuedThreadPool[OH-httpClient-nz-deconz-72aca4f3f1]@756c419e{STARTED,5<=5<=10,i=1,r=-1,q=0}[ReservedThreadExecutor@552afaf8{reserved=0/1,pending=0}] :

org.openhab.webclient:minThreadsShared = 10
org.openhab.webclient:maxThreadsShared = 120
org.openhab.webclient:minThreadsCustom = 10
org.openhab.webclient:maxThreadsCustom = 120

After searching a bit in this forum, I found several threads about memory problems, etc. This left me a bit unsure whether I should try to reduce openHAB’s memory usage on my current Raspi or instead buy another Raspi 4 with 8 GB RAM :man_shrugging:t2:

What am I using openHAB for?

  • Capture temperature data from 8 sensors via ZigBee
  • Control three LED lights via ZigBee

So here are my questions:

  1. Taking the above information into account, does 1 GB of memory seem like a reasonable high memory usage for you?
  2. Is there any way to reduce the number of threads without facing the aforementioned startup problem?
  3. Is it worth tuning openHAB’s memory usage (e. g. via JVM parameters) or would you rather recommend upgrading my Raspi to 8 GB?

Best regards and thx!

2

  1. No, that’s pretty typical on a 64-bit OH instance running in a container.

  2. I doubt it. OH is designed as an event driven multi threaded system.

  3. I recommend either:

  • don’t use Docker which does add some additional memory requirements
  • use 32-bit which has smaller memory requirements
  • moving to a machine with me RAM.

There is a new binding on the marketplace that will tell you in the logs if it looks like you’ve more heap allocated than you really need. That can reduce how much RAM OH grabs and holds onto.

The way Java works is it grabs a certain amount of memory when it starts. It doesn’t free that memory and give it back to the OS even if it’s not being used. However, when the max heap size is exceeded, you’ll get OutOfMemory exceptions in the log. This is an infusion that OH has a memory leak it a memory problem. Java manages it’s RAM usage within the heap.

You can change the minimum and maximum heap sizes by passing environment “JAVA_EXTRA_OPT= -Xms value.
-Xmx value” where “value” is replaced with the min and max heap size desired respectively.

But in general, 800 MB to 1.2 GB is pretty typical OH memory usage. OH is built to manage a while lot more than you are doing with it so it’s going to seem like it’s a lot. But if you had dozens to hundreds of devices, you’d find the memory requirements works remain roughly the same.

1 Like
3

Thank you very much for your in-depth explanation :muscle:t2: is it possible to run a 32 bit Docker image of openHAB on the 64 bit version of Raspberry Pi OS, is there a specific tag? Or would I have to “downgrade” the OS to 32 bit?

4

I don’t know. What really matters is that it’s the 32-bit version of Java running more than anything. If there is a version of the image that provides that it should work.