perf: batch marker-def fan-out to one PUT /nios/batch per compute

When a marker definition is created (or re-applied on data_link_type
change), the fan-out used to call inherit_marker -> update() on every
link, issuing one PUT /nio per link end (5000+ round-trips on a 2500-link
project). On started nodes each round-trip also reconfigured uBridge.

Two-phase fan-out:
- inherit_marker/start_marker gain memory_only: writes the marker into
  link._markers and refreshes _link_data without any HTTP/emit/dump.
- _apply_def_to_all_links applies memory-only to every link, then
  _batch_update_link_nios groups the updated NIO specs by compute and
  sends a single PUT /projects/{id}/nios/batch per compute.

compute: new PUT /projects/{id}/nios/batch endpoint with _get_existing_nio
+ _update_nio_binding dispatch (mirrors create_batch_nios), re-applies
filters+markers to uBridge on started nodes.

Precise per-marker operations (update_marker bpf change, stop_marker on
def delete) are untouched — they deliberately avoid a full reapply to
preserve sibling marker pcaps.
This commit is contained in:
YueGuobin 2026-08-11 08:44:57 +08:00
parent c877ed87a6
commit 48991f2d29
No known key found for this signature in database
4 changed files with 167 additions and 10 deletions

View File

@ -167,6 +167,65 @@ async def _add_nio_binding(node, adapter_number, port_number, nio):
)
def _get_existing_nio(node, adapter_number, port_number):
"""
Fetch the already-bound NIO for a port, preserving its UDP endpoints
(lport/rhost/rport) so a marker/filter update only changes markers/filters.
Dispatch keys off the manager class name, mirroring _add_nio_binding.
"""
manager_name = type(node.manager).__name__
if manager_name in ("Docker", "Qemu", "VMware", "VirtualBox"):
return node.get_nio(adapter_number)
elif manager_name == "IOU":
return node.get_nio(adapter_number, port_number)
elif manager_name in ("VPCS", "Builtin"):
return node.get_nio(port_number)
elif manager_name == "Dynamips":
# Dynamips routers expose NIOs via the slot/adapter; switches/hubs
# via get_nio(port).
if hasattr(node, "get_nio"):
import inspect as _inspect
if len(_inspect.signature(node.get_nio).parameters) >= 2:
return node.get_nio(adapter_number, port_number)
return node.get_nio(port_number)
raise HTTPException(
status_code=status.HTTP_400_BAD_REQUEST,
detail=f"Dynamips node '{node.name}' has no get_nio for batch update",
)
raise HTTPException(
status_code=status.HTTP_400_BAD_REQUEST,
detail=f"Batch NIO update not supported for node type '{manager_name}'",
)
async def _update_nio_binding(node, adapter_number, port_number, nio):
"""
Re-apply a NIO binding (filters + markers) to a started node's uBridge.
Dispatch keys off the manager class name, mirroring _add_nio_binding.
"""
manager_name = type(node.manager).__name__
if manager_name in ("Docker", "Qemu", "VMware", "VirtualBox"):
await node.adapter_update_nio_binding(adapter_number, nio)
elif manager_name == "IOU":
await node.adapter_update_nio_binding(adapter_number, port_number, nio)
elif manager_name == "VPCS":
await node.port_update_nio_binding(port_number, nio)
elif manager_name == "Dynamips":
if hasattr(node, "slot_update_nio_binding"):
await node.slot_update_nio_binding(adapter_number, port_number, nio)
else:
await node.update_nio(port_number, nio)
elif manager_name == "Builtin":
await node.update_nio(port_number, nio)
else:
raise HTTPException(
status_code=status.HTTP_400_BAD_REQUEST,
detail=f"Batch NIO update not supported for node type '{manager_name}'",
)
@router.post(
"/projects/{project_id}/nios/batch",
status_code=status.HTTP_201_CREATED,
@ -202,6 +261,37 @@ async def create_batch_nios(
return {"added": added}
@router.put(
"/projects/{project_id}/nios/batch",
status_code=status.HTTP_200_OK,
)
async def update_batch_nios(
project_id: UUID,
batch: schemas.BatchNIOCreate,
project: Project = Depends(dep_project),
) -> dict:
"""
Update many NIO bindings (filters + markers) across nodes in a single
request, re-applying them to uBridge on started nodes.
Used by the controller when a project-level marker definition changes and
must fan out to every affected link replacing one PUT /nio round-trip per
link end with one round-trip per compute. Each entry fetches the already-
bound NIO (preserving its UDP endpoints), overlays the new markers/filters,
and re-binds it.
"""
updated = 0
for entry in batch.nios:
node = project.get_node(entry.node_id)
nio = _get_existing_nio(node, entry.adapter_number, entry.port_number)
nio.filters = entry.nio.filters or {}
nio.markers = entry.nio.markers or {}
await _update_nio_binding(node, entry.adapter_number, entry.port_number, nio)
updated += 1
return {"updated": updated}
@router.get("/projects/{project_id}/files", response_model=List[schemas.ProjectFile])
async def get_compute_project_files(project: Project = Depends(dep_project)) -> List[schemas.ProjectFile]:
"""

View File

@ -114,7 +114,7 @@ class Link:
"""
return self._markers
async def inherit_marker(self, def_name, marker_def, dump=True):
async def inherit_marker(self, def_name, marker_def, dump=True, memory_only=False):
"""
Apply a project-level marker definition to this link.
@ -148,6 +148,7 @@ class Link:
enabled=not marker_def.get("paused", False),
inherited_from=def_name,
dump=dump,
memory_only=memory_only,
)
def _persist_markers(self):

View File

@ -1210,17 +1210,57 @@ class Project:
Fan out a single marker definition to every existing link in the project.
Links that have no capable node (``_MARKER_CAPABLE_TYPES``) are silently
skipped the marker can only live on a uBridge bridge.
Two-phase to avoid one HTTP round-trip per link end: (1) write the
inherited marker into each link's memory (``memory_only`` refreshes
``_link_data`` without pushing), then (2) batch-update every affected
NIO via a single ``PUT /projects/{id}/nios/batch`` per compute.
"""
d = self._marker_definitions[def_name]
# dump=False: per-link topology writes are the dominant cost on large
# projects — the callers (create/update_marker_definition) dump once
# after the fan-out.
await self._marker_apply_concurrently(
list(self._links.values()),
lambda link: link.inherit_marker(def_name, d, dump=False),
lambda link, e: f"Marker definition '{def_name}' could not be applied to link {link.id}: {e}",
)
affected = []
for link in self._links.values():
try:
await link.inherit_marker(def_name, d, dump=False, memory_only=True)
affected.append(link)
except ControllerError as e:
log.warning("Marker definition '%s' could not be applied to link %s: %s", def_name, link.id, e)
await self._batch_update_link_nios(affected)
async def _batch_update_link_nios(self, links):
"""
Push the current ``_link_data`` (markers/filters) of *links* to their
computes in one ``PUT /projects/{id}/nios/batch`` per compute replacing
one PUT /nio round-trip per link end. Started nodes re-apply uBridge;
stopped nodes update in memory.
"""
per_compute = {}
for link in links:
if len(link._link_data) < 2:
continue
for i, side in enumerate(link._nodes):
node = side["node"]
per_compute.setdefault(node.compute, []).append(
{
"node_id": node.id,
"adapter_number": side["adapter_number"],
"port_number": side["port_number"],
"nio": link._link_data[i],
}
)
async def _dispatch(compute, entries):
await compute.put(
f"/projects/{self._id}/nios/batch",
data={"nios": entries},
timeout=300,
)
if per_compute:
await asyncio.gather(
*[_dispatch(c, n) for c, n in per_compute.items()]
)
async def apply_defs_to_new_link(self, link):
"""

View File

@ -407,7 +407,7 @@ class UDPLink(Link):
# explicitly deletes a marker via the REST API, and a marker is torn
# down automatically only when its link is deleted.
async def start_marker(self, name, bpf, tag=None, direction=None, data_link_type="DLT_EN10MB", capture_node_id=None, color=None, highlight_duration=None, enabled=True, inherited_from=None, dump=True):
async def start_marker(self, name, bpf, tag=None, direction=None, data_link_type="DLT_EN10MB", capture_node_id=None, color=None, highlight_duration=None, enabled=True, inherited_from=None, dump=True, memory_only=False):
"""
Attach a traffic-insight marker to this link.
@ -463,6 +463,12 @@ class UDPLink(Link):
if inherited_from:
marker_entry["inherited_from"] = inherited_from
self._markers[name] = marker_entry
if memory_only:
# Project-open prepare / marker-def fan-out: only refresh the
# in-memory NIO specs so a later batch dispatch carries the new
# markers — no per-link update HTTP, notification or dump.
self._refresh_link_data()
return
if self._created:
await self.update()
self._project.emit_notification("link.updated", self.asdict())
@ -471,6 +477,26 @@ class UDPLink(Link):
if dump:
self._project.dump()
def _refresh_link_data(self):
"""
Recompute the filters / markers / suspend fields of ``_link_data`` from
the current link state without pushing to computes. Used by the
memory-only marker path so a batch dispatch picks up the new markers.
"""
if len(self._link_data) < 2:
return
node1 = self._nodes[0]["node"]
node2 = self._nodes[1]["node"]
node1_filters, node2_filters = self._get_node_filters(node1, node2)
node1_markers, node2_markers = self._get_node_markers(node1, node2)
self._link_data[0]["filters"] = node1_filters
self._link_data[0]["markers"] = node1_markers
self._link_data[0]["suspend"] = self._suspended
self._link_data[1]["filters"] = node2_filters
self._link_data[1]["markers"] = node2_markers
self._link_data[1]["suspend"] = self._suspended
async def stop_marker(self, name, inherited=False, dump=True):
"""
Remove a traffic-insight marker from this link.