If you're running docker with devicemapper as the storage driver, there are a few steps to recover after deleting
$ sudo service docker stop
$ sudo rm -rf /var/lib/docker
$ sudo lvremove docker
$ sudo service docker start
This will help you recover from errors like
Error starting daemon: error initializing graphdriver: devicemapper: Non existing device docker-docker–pool
Error starting daemon: error initializing graphdriver: Unable to take ownership of thin-pool (dockerVG-docker--pool) that already has used data blocks
Thanks to hints from redhat bugzilla and docker forums.
This is part of our antipatterns series. Ouch!
With the recent bugs in ecs-agent 1.8.0, you may be trying to roll back to amzn-ami-2015.09.e or earlier to get a last-known-good ecs agent.
If you have
yum update in your userdata, however, it updates
ecs-init and that, in turn, will auto-upgrade you to 1.8.0—rolling back to an older image won't help!
But... you don't want to get rid of
yum update it from your userdata because of fun CVEs in glibc and openssh.
yum update --exclude=ecs-init
Best of both worlds: you get the latest security patches and you can roll back to whatever agent you want!
Confirmed to work with Julien of AWS Support. Thanks Julien!