- Dockerfile is very much like a makefile creating the same instance each time
- with enough nodes the swarm has some survivability
- the docker networks can be encrypted for additional security
- the docker networks can be segmented stitching the systems that are permitted to communicate
- when combined with traefik there is some dynamic deploy that I like including let's encrypt and SSL
- in recent history it has been reported that there are some bad actors creating fake containers and there is no curated container services that are not stupid expensive. This is a common problem for open source.
- deploying docker services and stacks can relocate them anywhere in the swarm, however, if you use persistent volumes they do not follow and so you need a distributed filesystem, NAS, or SAN. All of which have their own risks and costs.
- repairing a damaged cluster means rebuilding it all. This is typical but seriously tricky to be consistent as well as keeping the docs up to date. For example I had to push my swarm source outside of my network so I could deploy it differently if there was a major failure in the lab.
Right now the network filesystem is a problem without a solution.