On 12 February 2017 at 16:08, Johannes Truschnigg <
johannes@truschnigg.info> wrote:
> Sure you CAN run an embarassingly stateful system like a RDBMS on a
> containerization platform designed to run stateless,
Running (Docker) containers does not necessitate running stateless, many
properties of Docker are very useful for stateful systems, I'm thinking of
- isolation between environments
- deploying tested and verified Docker images
- allowing your developers to use the same Docker image as production
> the pieces of the puzzle that are storage drivers (AUFS and friends) and
> enable container networking, which is not unheard of.
An example of a pattern that avoids these issues, but does give you the
isolation and deployment benefits is:
- run 1 container on 1 node
- use bind mounts
- use --net=host
e.g.: docker run -ti -v /pgdata/abc:/var/lib/postgresql/data --net=host postgres
I'm still on the fence whether or not running PostgreSQL inside containers is
advisable or not, I've actively ran many instances successfully without any
major issues. I'm currently thinking that if you run a lot of small databases
for many teams (a Database as a Service) it would be valueable. If you however
want to squeeze all the performance out of your single large database it might
not be the best way of running your system.
regards,
Feike