urg_node failed on Intel NUC / swapping issue?

asked 2017-11-20 04:45:32 -0500

fabriceN gravatar image

updated 2017-11-27 04:45:31 -0500

I am facing a problem running urg_node with an Hokuyo 10LX on an embedded computer.

Here is the environment, I am running ROS kinetics on Ubuntu 16.06LTS on an INTEL NUC 5ITRYH with 8GO DDR4 and SSD 200Go.

When I start urg_node, it works perfectly but as soon as I open a terminal for instance, I got the following warning:

process[urg_node-2]: started with pid [5099] 
[ INFO] [1511171038.966908047]: Connected to network device with ID: H1642740 
[ INFO] [1511171038.979387775]: Starting calibration. This will take a few seconds.
[ WARN] [1511171038.979419672]: Time calibration is still experimental. 
[ INFO] [1511171039.982890964]: Calibration finished. Latency is: -0.0208. 
[ INFO] [1511171039.998174108]: Streaming data.
[ WARN] [1511171050.027225044]: Could not grab single echo scan.
[ERROR] [1511171051.692534732]: Error count exceeded limit, reconnecting. 
[ INFO] [1511171053.704590758]: Connected to network device with ID: H1642740 
[ INFO] [1511171053.716937334]: Starting calibration. This will take a few seconds.
[ WARN] [1511171053.717063649]: Time calibration is still experimental. 
[ INFO] [1511171054.718481924]: Calibration finished. Latency is: -0.0212. 
[ INFO] [1511171054.734664754]: Streaming data.

And it if I start a slam with RVIZ, i finally ended up with:

[ WARN] [1511170394.929138196]: Time calibration is still experimental.
[FATAL] [1511170395.851344341]: Could not calibrate time offset: Cannot get scan to measure time stamp offset. [urg_node-2] process has died [pid 4389, exit code -5, cmd /home/fabrice/catkin_demo_lidar_ws/devel/lib/urg_node/urg_node __name:=urg_node __log:=/home/fabrice/.ros/log/c61d770e-cdd5-11e7-b1f1-94c691132292/urg_node-2.log]. log file: /home/fabrice/.ros/log/c61d770e-cdd5-11e7-b1f1-94c691132292/urg_node-2*.log

It looks like, at first glance, that the NUC takes too much time to swap between tasks but I am I have to admit that I am surprise.

I tried to change the priority between tasks with renice, but without success.

The same programs (urg_node, hector slam...) work perfectly on my laptop with ROS kinetics/Ubunto 16.04LTS

Does anybody had the same experience and can share how to handle it?

Thank you very much

Fabrice

edit retag flag offensive close merge delete

Comments

You'll need to provide enough information to reproduce the problem to get help. If you're boardering on overloading a machine starting a new set of processes like visualization and SLAM I would not be surprised to see unexpected behaviors. Have you checked that you're not getting OOM killed?

tfoote gravatar imagetfoote ( 2017-11-26 15:59:33 -0500 )edit

@tfoote thank you for your message, actually after searching on the net, it looks like that it might be an issue with the Intel LAN driver. And so far the workaround is to go thru a USB to Etehrnet cable and everything is working fine now. Does anyone out there experienced the same problem?

fabriceN gravatar imagefabriceN ( 2017-11-27 04:49:13 -0500 )edit