Skip to content

op-acceptance-tests: cover an op-reth restart with persisted state under a running op-node #23597

Closed

Description

@sebastianst

Part of #21744.

Gap

The op-reth-test-engine stack (#21787) deletes RestartOpGeth, the action test that restarted the in-process op-geth EL over its persistent datadir while op-node kept running. The ephemeral test engine cannot model that: its only restart semantics is total state loss, covered by RestartEngineFreshState. The closest sysgo test, TestNotTruncateDatabaseOnRestartWithExistingDatabase, restarts op-node, not the EL. So nothing tests an EL that comes back with its chain intact under a running op-node, the most common restart in production.

Work

Add a sysgo acceptance test: stop op-reth under a running op-node (L2ELNode.Stop/Start; op-reth's sysgo datadir persists across the restart), start it again, and assert that op-node did not reset and that unsafe and safe heads advance again. This is the system-level home for that behaviour (real EL + op-node's engine-RPC retry path), so it is not part of the test-engine stack.

🤖 Co-created with Claude Fable 5.1

Metadata

Metadata

Assignees

Labels

No labels
No labels

Type

No type

Projects

No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions