Containers overlapping when swarm-sync service updated

[kevb]

When swarm-sync itself is updated, it receives a SIGTERM and will wait until the current update cycle is complete to exit.

The service is by default deployed with stop-first update strategy meaning the new container for the updated swarm-sync should not start until the earlier one has completed.

However, I am seeing 2 containers running at the same time:

e1055a821402        swarmpack/swarm-sync:v0.1.43                       "./start.sh"             About a minute ago   Up 10 seconds       80/tcp                    tools_swarm-sync.1.y3o84m5je7b1ce3vhitl879nh
a3a96b39469c        swarmpack/swarm-sync:v0.1.45                       "./start.sh"             About a minute ago   Up About a minute   80/tcp                    tools_swarm-sync.1.s72a0tkgwusaml54u7q7jodf0

Both containers are up and this causes a big problem because the state of the earlier one has not been saved yet, and the new one starts to do the same tasks causing a repeating loop of swarm-sync updates which eventually resolve by chance of one completing fast enough to prevent this.

I will try to create a demo project to explore the issue further...