feat(raw): re-migrate discovered-node injection onto the raw device subtree — IRediscoverable is inert in v3
#507
Assignees
dohertj2 (dohertj2)
Clear assignees
dohertj2 (dohertj2)
No Assignees
Notifications
Due Date
No due date set.
Dependencies
No dependencies set.
Reference: dohertj2/lmxopcua#507
Reference in New Issue
Block a user
Delete Branch "%!s()"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
IRediscoverable.OnRediscoveryNeededcurrently changes nothing.DriverHostActor.HandleDiscoveredNodes(
src/Server/ZB.MOM.WW.OtOpcUa.Runtime/Drivers/DriverHostActor.cs:986) hard-returns after a Debug log — adeliberate v3 Batch-4 guard, whose own comment ends "Re-migrating injection onto the raw device subtree is a
separate follow-up." This is that follow-up.
Why it was short-circuited. In v2 a driver-connected
FixedTreewas grafted under an equipment folder(equipment bound a driver). v3 retired that binding — equipment references raw tags via
UnsTagReference, anddiscovered raw tags are authored explicitly through the Batch-2
/rawbrowse-commit flow. The downstreammapper/materialiser were only half-migrated to the Raw tree and are still rooted at an equipment id, so letting
the message through would materialise incoherent nodes. Short-circuiting was correct; leaving it permanent is
the gap.
Effect today. Every driver implementing
IRediscoverablefires into a void:DeployWatcherobserving a Galaxy redeploy@tagschangeThe operator-visible symptom is uniform: the edge gains a point and the OPC UA tree does not change until
someone redeploys. The driver-side decision is log-observable (each driver logs that it raised rediscovery),
so this is a platform gap, not a per-driver bug — fixing it once lights up all four.
Scope sketch. Re-root the discovered-node mapper/materialiser at a Device in the
/rawtree ratherthan at an equipment id, and decide the authoring contract: auto-materialise discovered leaves, or surface them
as a pending-review set the operator commits (the latter matches the existing
/rawbrowse-commit flow andavoids a chatty edge rewriting the address space unattended). Whichever is chosen,
_discoveredByDriverandthe redeploy re-inject tail need to come back with it.
Acceptance. A driver raising
OnRediscoveryNeededresults in the new leaf being reachable in the servedaddress space (auto, or after an explicit commit) without a redeploy, live-verified on docker-dev against the
Sparkplug simulator (
docker restart otopcua-sparkplug-simforces a fresh birth; births are never retained).CLOSED AS SUPERSEDED — the injection path was deleted, not migrated (
adce27e7, merged914d70bf).Why the requested migration was not done
This issue asks to re-migrate discovered-node injection onto the raw device subtree. Reading the retained code showed there was nothing to migrate: its two inputs are structurally empty in v3.
EquipmentNode.DriverInstanceId—AddressSpaceComposer.cs:468-470, 514-517EquipmentTags—DeploymentArtifact.cs:614So removing the
#pragma warning disable CS0162guard would have changed a log line and injected nothing. The deferment register had recorded this as a "deferred (deliberate guard)" — too generous; it was dead code behind a guard that made it look revivable.Deleted
HandleDiscoveredNodes,PartitionDiscoveredByDeviceHost,ApplyDiscoveredPlansForDriver, the redeploy re-inject tail,DiscoveredNodeMapper,AddressSpaceApplier.MaterialiseDiscoveredNodes,OpcUaPublishActor.MaterialiseDiscoveredNodes, the connect-time discovery loop (StartDiscovery/RediscoverTick/DiscoveredNodesReady/TriggerRediscovery), and 4 files' worth of tests.The connect-time loop went with it because it had no remaining consumer — it was browsing real devices up to ~15× per connect and dropping the result.
What replaces it
Discovered raw tags are authored through the
/rawbrowse-commit flow, which is the deliberate v3 posture: nodes exist because an operator approved them and a deployment artifact records them.ITagDiscoveryitself stays — the browse picker drives it throughCommons/Browsing/DiscoveryDriverBrowser, which never went through the actor.IRediscoverablenow surfaces a re-browse prompt on/hostsinstead (#518).One correction to the record
The 18
DiscoveryInjectionDormantV3skipped tests were described as "blocked on #507". They were v2 characterization tests asserting an equipment-rooted graft (EquipmentRootNodeId == "EQ-1") that v3 cannot produce — they could never have unskipped as written, with or without this issue. Deleted.Runtime.Testsskipped went 31 → 13.