Ghost Drive

[bkirshh]

On occasion when the power is abruptly cut to the RaspberryPi running dencam, USB attached storage drives (microSD cards in card readers) do not appear to be properly ejected. The name of the drive appears to stick in /media/pi.

Several of my Pi's (I am currently testing five with four different batteries) have been neglecting to display the second status page. This, as I have come to find out, is because the system is remembering the names of my attached storage when the pi is not properly shut down (nearly every time I just pull the plug). This is easy enough to solve (just ssh in and cd /media/pi and sudo rmdir X). But in the field, this could be catastrophic as these pseudo-replicated drives prevent dencam.py from writing to either the external or internal storage.

Potential fix:

EDIA_DIR = '/media/pi'

def clearGhostDrives():
    """Delete non-existent USB drives"""
    # When power is removed (no shut-down) and a USB drive is removed, 
    #  then the power is restored, a ghost folder is left in /media/pi/
    # Distinguishing between ghost folders and real ones has proven to be too much of a challenge
    # Instead, this function deletes all the USB folders
    # This is executed when powering up, before the USB drive is mounted
    # Therefore, after this removes all the folders, the system mounts the USB drive that actually exists

    # Locals
    DEL_DIR_CMD = 'sudo rm -r %s/%s'

    usbDrivesList = os.listdir(MEDIA_DIR)
    if len(usbDrivesList) > 1:
        for usbDriveName in usbDrivesList:
            doLinuxCmd(DEL_DIR_CMD % (MEDIA_DIR, usbDriveName)) 

[iamgarcia]

@bkirshh I have been unsuccessful in being able to replicate this issue. Are you able to replicate this issue? If so, could you provide me with the exact steps that you did to cause the ghost drive issue? Thanks!

[bkirshh]

After my devices stabilize (we fix the slow DenCam issue) I'll try again to replicate. In the meantime can you try pulling the power on a functioning DenCam over and over again, maybe ten times. Don't properly shut it down, just pull the power as the files are writing. See what happens.

[iingram]

There was recent evidence that this issue was rearing its ugly head. Alex's fix from way back when that he was unable to truly test because the issue couldn't be recreated was cherry-picked and tested to the degree it could be and incorporated into v0.0.7 because it was deemed the risks incurred by not doing so outweighed the risks incurred by doing so. We should try to do some more testing to both verify the fix really addresses all the ways the problem can manifest and also that there is no regression.