Yup, connected to RPi3. Though I compile everything myself, I never use packages. Maybe that makes a difference? I don't know. I also flashed the camera with the ASI120MC-compatible firmware.
Just tried another thing which helped on the crashes, it seems. I previously started indi as an iniscript containing
sudo -u pi bash -c 'indiserver -f /home/pi/indiFIFO &>/home/pi/indi.log' &
sudo -u pi echo start indi_asi_ccd >/home/pi/indiFIFO
and similar lines to start the other drivers. This should run everthing as user pi, not as root and "ps xauf" also confirmed this (it seems!).
Then I tried to change to running it as
/sbin/runuser -u pi "/home/pi/startindi.sh"
where the startindi.sh contains the above lines without the sudo.
Everything, including heavy ssh file transfer, have been working ever since I changed this and I will keep my fingers crossed
I now have tried to change the SD card of the RPi (the old was too small anyway...). I did a fresh install og jessie and indi. The camera worked fine all night (YEAH!), but when the sun came up, the exposures got shorter (but still only one exposure pr. minute) problems started to happen and finally the system crashed and became unreachable. A few minutes after the exposures started to fail consistently, I get these in /var/log/messages
/var/log/messages:Dec 6 08:10:51 observatory kernel: [39423.368478] usb 1-1.4.1: reset high-speed USB device number 6 using dwc_otg
Device 6 is the ASI120MC.
After the camera stopped, the capture script reportet problems like
No ZWO CCD ASI120MC.CCD_CONTROLS.Gain from localhost:7624
Seems the asi driver as not working correctly for some reason.
One more odd thing... An hour or so before the crash I noticed that the indi_qhy_ccd was using about 13% of the memory of the RPi. Now I just restarted the RPi and indi, and 'top' says the qhy driver uses 0.6% of the memory. I did not use the qhy driver yesterday (just started it), as it is for my main CCD on the telescope. I will monitor the memory usage of indi to see if there are any memory leaks...
There do seem to be some memory leak in the qhy driver. 10 minutes later (without having my qhy even connected) the memory usage was at 2.6%. It does not, however, seem to have anything to do with my problems as everything still crashes without loading the qhy driver...
I wrote a small script to test this, it is attached here. It captures 100 exposures of 1s, 0.1s, 0.01s, 0.001s and 0.0001s each.
It successfully does 100 exposures of 1s, 0.1s and 0.01s. For exposures of 0.01s some frames are not captured and the following type of lines starts appear in /var/log/messages
Dec 6 15:44:22 observatory kernel: [18577.623140] usb 1-1.4.1: reset high-speed USB device number 6 using dwc_otg
When it reaches exposures of 0.001s, the RPi crashes after 5 exposures. The last lines of /var/log/messages:
Dec 6 15:45:18 observatory kernel: [18633.303506] usb 1-1.4.1: reset high-speed USB device number 6 using dwc_otg
Dec 6 15:45:20 observatory kernel: [18635.433524] usb 1-1.4.1: reset high-speed USB device number 6 using dwc_otg
Dec 6 15:45:21 observatory kernel: [18635.855286] usb 1-1.4.1: USB disconnect, device number 6
I will try to limit my capture script to >1s exposures to see if this is a usable work-a-round...
The precompiled indi for RPi also crashed with long exposures, not only short. It just takes longer for it to crash...
I tried to compile an old version from git ( f6f260b4c70f93ed4d32d488a9a65a372af56b11 ) and it have worked fine for a couple of hours with both long and short exposures. It can not, however, go below 0.001s as the precompiled version can. I am keeping my fingers crossed that this is now fixed
By disconnecting all other USB pheripherials and connecting the ASI120 directly to the RPi's USB ports and limiting exposures to >0.5s, I have now managed to get the system running for 2 days without a crash. If I do short exposures with the ASI connected like this, the entire system does not crash but the exposures does not work and indi or the indi_asi_ccd driver seems to crash. Major step forward, but still not great!
Apparently the cameras are not campatible with mac and linux, see zwoug.org/viewtopic.php?f=17&t=6668&p=11652#p11640
"the hardware performance of ASI USB2.0 cameras is not very powerful
so linux and Mac is not supported very well"
This ought to be put in a FAQ somewhere to prevent others from wasting money on these cameras. The cheapest solution for me seems to be tu buy a dedicated RPi for my all sky since it now seems to work with the longer exposures when nothing else is connected.