ii hailo-all 4.20.0 all Hailo support (metapackage)
ii hailo-tappas-core 3.31.0+1-1 arm64 This package contains the core components of the Hailo Tappas platform.
ii hailofw 4.20.0-1 all Hailo firmware
ii hailort 4.20.0-1 arm64 HailoRT
ii python3-hailort 4.20.0-1 arm64 HailoRT Python API, which wraps the runtime library
ii rpicam-apps-hailo-postprocess 1.7.0-1 arm64 rpicam-apps-hailo
The application, at the moment, grabs a frame from a camera and then runs it through pose estimation before taking the results and doing some custom processing with them. It runs fine but periodically the app just exits without warning. The appâs logs donât show any errors, just a sudden stop. When I look in the hailort.log file I see:
Can you guide me how you install the libraries, hailo packages as in the picture.
i am downloading rpi5-exampes but it is giving error when ./install.sh .
Installing missing pip packages⊠Installing âhailortâ via helper script
â HAILORT_VERSION = 4.20.0
â TAPPAS_CORE_VERSION= 3.31.0
â DOWNLOAD_DIR = hailo_temp_resources
â install Tapas? = false
â install HailoRT? = true
when i check dpkg -l | grep hailo
±-----------------------±-----------±-------±--------------------------------------------------------------+
| Package Name | Version | Arch | Description |
±-----------------------±-----------±-------±--------------------------------------------------------------+
| hailo-tappas-core | 3.31.0+1-1 | arm64 | Core components of the Hailo Tappas platform |
| hailort | 4.22.0 | arm64 | HailoRT runtime |
| hailort-pcie-driver | 4.22.0 | all | Hailo PCIe driver and firmware |
using only sudo apt install hailo-allâ. ?
Have you tried downloading rpi5-example? Did you follow the github instructions or is there anything else you need to keep in mind, like about venv?
Is this still an issue ?
Looking at those two log lines:
[event.cpp:113] [eventfd_poll] poll failed with errno=4
[event.cpp:55] [wait] CHECK_SUCCESS failed with status=HAILO_INTERNAL_FAILURE(8)
Hereâs whatâs happening:
The first line shows errno=4, which corresponds to EINTR - meaning the Linux poll() system call got interrupted by a signal. The second line shows HailoRT converting this into a HAILO_INTERNAL_FAILURE status (code 8), which is basically its catch-all for âsomething unexpected went wrong.â
The issue is that HailoRTâs event polling doesnât automatically retry when this happens - it just treats any error as an internal failure.
How to fix it:
Add retry logic - Catch HAILO_INTERNAL_FAILURE and retry the operation.
I can see youâve got both HailoRT version 4.22 and 4.20 installed on your system, which is causing some confusion - the system doesnât know which one to use.
Iâd suggest sticking with version 4.20 for now and removing 4.22. Version 4.22 will be officially released with hailo-all soon, so itâs better to wait for that proper release rather than mixing versions.
This should clear up any conflicts youâre experiencing!
I will definitely look into adding a catch for HAILO_INTERNAL_FAILURE. I wasnât aware you could catch that, but thatâs good to know.
In the meanwhile I think I was able to resolve the problem. When I first installed the Hailo board a few months back I remember the instructions listing the PCIe Gen 3 setting as optional, but I see itâs now not marked optional. I tried enabling it and since then my unit has been fine, so I think upgrading to the newest kernel AND enabling PCIe Gen 3 was the fix for me.
Is HAILO_INTERNAL_FAILURE an exception that can be caught in python, or only c++? My app is python.
One of my devices still occasionally has this issue, but im in the process of starting clean with that device to see if it clears up. Either way Iâd like to be able to catch and handle that error in python if it does occur.
Are we sure the HAILO_INTERNAL_FAILURE is being raised as a python exception? I have a try except around the code that looks like this:
try: hailo_inference.run() preprocess.join() # To signal processing process to exit output_queue.put(None) postprocess.join()
check_process_errors(preprocess, postprocess)
except Exception as e:
logger.error(f"Inference error: {e}")
# Ensure cleanup if there's an error
input_queue.close()
output_queue.close()
preprocess.terminate()
postprocess.terminate()
os._exit(1) # Force exit on error
Shouldnât the general Exception clause catch the HailoRTException? That seems to imply itâs not being raised up to Python so it canât be caught by the app. Is that correct?
HAILO_INTERNAL_FAILURE is a C++ status code from libhailort. The Python wrapper only converts it to a HailoRTException when the error crosses the Python-C++ boundary. In your case, the error occurs inside the C++ pipeline thread and only gets logged - it never propagates back to your Python code.
Your current code:
try:
hailo_inference.run() # starts async pipeline
preprocess.join()
postprocess.join()
except Exception as e:
# This never executes because the error stays in C++
...
How to fix it
Option 1: Use synchronous API
try:
configured_model.run([bindings], timeout_ms)
except HailoRTException as e:
logger.error(f"Hailo error: {e}")
# Handle the error properly
Option 2: Use async with callback
def on_complete(completion_info, _):
if completion_info.exception:
logger.error(f"Inference failed: {completion_info.exception}")
# Handle the error
job = configured_model.run_async([bindings], callback=on_complete)
Yes, HailoRTException is the Python exception class for internal statuses (including HAILO_INTERNAL_FAILURE).
No, your general except Exception wonât catch it unless you force the status back into Pythonâeither by using the sync API or by explicitly checking completion_info.exception in an async callback.
how you install the libraries, hailo packages as in the picture.
Installing missing pip packagesâŠ