ConceptsDisks, scaling & suspend
Deploys with a disk
A service with a disk has one instance. A deploy stops the old instance before it starts the new one.
The one instance of the service uses the disk.
1 / 5
Two containers must not write to one disk at the same time. Thus a service with a disk has exactly 1 instance. Ferry refuses to scale it higher.
Each deploy and each restart stops the old instance, then starts the new instance. There is no zero downtime for this service.
In the deploy log
==> Mounting disk at /data
==> Limits: 512 MiB memory, 1 CPU per instance
==> Stopping 1 old instance(s) first (services with a disk are redeployed in place)
==> Starting 1 instance(s)A service with no disk does the opposite: the new instance starts first. See The steps of a deploy.