docker swarm - global or replicated?

[madhums]

Hi, I am using this library to create a self-managed docker swarm mongodb service.

I have a question: should the backup be run only on the managed node? And should the service be global instead of replicated? Because if this is replicated or is run on multiple nodes, we risk the backup being run simultaneously and multiple times.

docker service create --name mgob \
  -p 8090:8090 \
  --mount type=bind,src=./mgob/config,dst=/config \
  --mount type=bind,src=./mgob/storage,dst=/storage \
  -e BACKUP__TARGET_URI=<insert-connection-string-here> \
  -e BACKUP__S3_URL=<set-values> \
  -e BACKUP__S3_BUCKET=<set-values> \
  -e BACKUP__S3_ACCESS_KEY=<set-values> \
  -e BACKUP__S3_SECRET_KEY=<set-values> \
  # --constraint node.labels.mongo.replica=1 \ # to restrict it to be run only on single manager node
  # --mode global \ # should it be global or replicated?
  maxisam/mgob:1.10.14-aws -LogLevel=info

And the reason I'd like to create this as a service is to ensure fault-tolerance. If in case a manager node fails, it's replicated in another manager node. Does this make sense?

[madhums]

Ok, this should not be global because otherwise it will be deployed on every node and we don't want the backup to be run on all nodes.