Operate4 min read

Upgrading running processes: migration and batch operations

Move live instances onto a new version of a process, and cancel, retry or update many instances in one go.

In short

01
New instances use the new version; old ones stay put until you migrate them.
02
You map each old step to its new step.
03
Batches report a result per instance and never fail as a whole.

Deploying a new version of a process does not touch instances that are already running. They finish on the version they started with. Migration is how you move them across when you need to.

Migrating instances

Deploy the new version

It must have the same process key as the old one.

Map the steps

For each step an instance might be waiting at, say which step it becomes in the new version.

Pick the instances

Choose which running instances to move.

Run and review

Each instance comes back as migrated, or not migrated with the reason.

Which instances can move

Instance state	Can migrate?
Waiting at a mapped step, on a single path	Yes
Waiting at a step you did not map	No — add it to the mapping
In the middle of parallel paths	No — wait until the paths join
Finished, failed or not waiting	No

An instance that cannot move is reported with its reason and left untouched. The others in the same run still migrate.

Batch operations

Migration is one of five operations you can run across many targets at once.

Batch type	Applies to	Does
Cancel	Process instances	Terminates them
Retry	Jobs	Makes failed jobs available again
Resolve incident	Incidents	Retries the failed step of each
Migrate	Process instances	Moves them to another version
Set variables	Process instances	Writes the same variables to each

A batch runs in the background. Its progress shows how many items are done and how many failed, and each failed item carries its own reason. A batch that is interrupted by a restart picks up where it left off.

Pending

Running

Completed

or Completed with errors