The image compiles OK, it downloads to the device, the device comes online, I see the logs scrolling and I can get in remotely using the terminal. However, when start.sh runs it does a
docker compose up
and at this point, I get errors saying that the docker index cannot be reached and everything stops.
Just as importantly, the device stops updating its image: i.e. if I update the dockerfile and push a new version, it gets rebuilt on resin.io but does not get downloaded to the device. The only solution is to delete the device and start afresh.
By installing some test tools I have established that the problem seems to be with name resolution. I can ping IP addresses but not dns names.
I’ve tried force pushing a different image and this works fine, so it’s not the network or the device hardware.
There’s very little in the compose file, other than comments. here’s the bit that isn’t commented out:
mysql:
image: "hypriot/rpi-mysql"
environment:
# set mysql credentials
- MYSQL_ROOT_PASSWORD=secure
- MYSQL_USER=flight
- MYSQL_PASSWORD=flight
- MYSQL_DATABASE=flight
volumes:
# bind the folder /opt/flight/mysql on the raspberry as datastorage
- /opt/flight/mysql:/var/lib/mysql
Thanks for the UUID. The only thing of note I can see in the compose file is that of binding /opt/flight/mysql to the MySQL database. This isn’t a persistent storage area, so will vanish the next time a reboot/app restart occurs (and this might be causing your error, though without the actual error message I can only speculate).
This shouldn’t be preventing a new app build from being downloaded to your device, however. I shall try and replicate this issue and get back to you.
I’ve not been able to replicate this. If you could supply an archive with a cut-down example where this occurs, or a Github repo of the same, this would help us try and replicate the issue.
Just in case you missed it, we’d like to let you know that resin.io now natively supports multicontainer applications. The docker-in-docker workaroud is no longer needed, and the DNS issues mentioned in this thread should by now be resolved.