My BBW worked for a long time with UPS and an "industrial" SD card (which
in theory has wear leveling and asynch shutdown tolerance) but now that card
has gone unbootable.
No level of SD write that should exhaust what flash is in theory capable of,
but theory and practice...
Any success stories out there using bb in unattended situations?
Are the odds perhaps better if you never write the SD or eMMC at all, and if so
is there a distribution set up to work this way (ramdisk default or the like)?
Thanks,
Britton
I have an xM that's been running outside taking webcam shots for over
3 years, same sd card..
It's running Debian with a 'ro', (read only) root file system... RAM
used as swap...
Regards,
I use two BBBs in remote locations, running Ubuntu from eMMC. They are networked (I can reach them) and if really necessary I could get someone to power cycle them the next day, so not sure how far that counts as unattended. Remote hands so far have not been necessary though. Have seen uptimes over 3 months.
They do write logs (icinga/nagios, cacti, syslog for appliance) but don’t get rebooted very often so not sure how much this says about the likelihood of eMMC corruption issues. I would expect ext4 to catch most of these though?
The limited instability I have seen with BBBs is where wireless USB devices are used, but even one with two USB wifi adapters has seen uptimes over 50 days.
I have yet to experience an eMMC or SD that becomes spontaneously unbootable.
Best,
-Bert
While not really unattended, I have had a BBB running Debian 7 ( wheezy ) for as long as as 45 days with no reboots. I have accessed it via SSH, and did some testing on it during this time.
A couple of differences here besides the obvious is that I have it powered via USB ( from a Windows 7 laptop ), and the Debian used is a network boot configuration. e.g. it uses no emmc or flash media in this configuration.
Anyhow, the BBB in this situation would need some form of UPS, as when coming back up form a reboot( power blip would do the same thing ) sometimes it can hang. Either that or a remote “switch” for resetting the power on the device. Also as Robert mentioned above read only media would be a requirement if you expect to get any reasonable life time out of the device. However, the BBB can also boot via a USB device if that is also an option. I’m thinking hard drive here for longevity but . . .
One thing we do running unattended BBBs is to have the watchdog package installed and tied to the hw watchdog on the AM335x, and then ensure after a reboot that you application and any networking return to their previous state (before reboot). This way even occasional unexpected crashes will generally cause a reboot and give you a downtime of no more than a couple of minutes.
We implemented this when we were getting fairly regular problems, but using the new 3.12 kernel and Debian Wheezy (and various other tweaks to our application), we have not experienced any watchdog reboots for a number of days, and have been transferring hundreds/thousands of gigabytes over usb wifi - but we’ll leave it in as a backup nonetheless.