EFS Mount Filesystem: connection reset by peer

[richardcalahan]

I've seen the same issue raised in various forums related to the creation of an EFS system on Beanstalk.

Including both the create efs and mount efs extensions in a single (initial) deployment consistently fails in the 'mount efs' script with:

mount.nfs4: Connection reset by peer

Various people point out that it's an issue with the mount target and ec2 instances not sharing security groups, or at least not having granted the proper access to each others' security groups.

This is probably true, but after failure I've logged into the ec2 instance that attempted to run the 'mount efs' script, ran it manually from the /tmp directory, and it always succeeds. The 'mount efs' script just runs too soon (at least on Linux 2).

I simply changed the call from "commands" to "container_commands" here. At this point all initial deployments, including rebuilds, succeed. Seems it just need to mount a little later.

I suppose one could also run the mount-efs.sh as a predeploy platform hook. In theory that would happen around the same time.

Anyhow, perhaps the docs should be updated? (Or maybe im just missing it)