goodboy
/
tractor
1
0
Fork 1
Code Issues 1 Pull Requests 4 Packages Projects Releases Wiki Activity

Remote `Context` cancellation semantics rework B) 

This adds remote cancellation semantics to our `tractor.Context`
machinery to more closely match that of `trio.CancelScope` but
with operational differences to handle the nature of parallel tasks interoperating
across multiple memory boundaries:

- if an actor task cancels some context it has opened via
  `Context.cancel()`, the remote (scope linked) task will be cancelled
  using the normal `CancelScope` semantics of `trio` meaning the remote
  cancel scope surrounding the far side task is cancelled and
  `trio.Cancelled`s are expected to be raised in that scope as per
  normal `trio` operation, and in the case where no error is raised
  in that remote scope, a `ContextCancelled` error is raised inside the
  runtime machinery and relayed back to the opener/caller side of the
  context.
- if any actor task cancels a full remote actor runtime using
  `Portal.cancel_actor()` the same semantics as above apply except every
  other remote actor task which also has an open context with the actor
  which was cancelled will also be sent a `ContextCancelled` **but**
  with the `.canceller` field set to the uid of the original cancel
  requesting actor.

This changeset also includes a more "proper" solution to the issue of
"allowing overruns" during streaming without attempting to implement any
form of IPC streaming backpressure. Implementing task-granularity
backpressure cross-process turns out to be more or less impossible
without augmenting out streaming protocol (likely at the cost of
performance). Further allowing overruns requires special care since
any blocking of the runtime RPC msg loop task effectively can block
control msgs such as cancels and stream terminations.

The implementation details per abstraction layer are as follows.

._streaming.Context:
- add a new contructor factor func `mk_context()` which provides
  a strictly private init-er whilst allowing us to not have to define
  an `.__init__()` on the type def.
- add public `.cancel_called` and `.cancel_called_remote` properties.
- general rename of what was the internal `._backpressure` var to
  `._allow_overruns: bool`.
- move the old contents of `Actor._push_result()` into a new
  `._deliver_msg()` allowing for better encapsulation of per-ctx
  msg handling.
 - always check for received 'error' msgs and process them with the new
   `_maybe_cancel_and_set_remote_error()` **before** any msg delivery to
   the local task, thus guaranteeing error and cancellation handling
   despite any overflow handling.
- add a new `._drain_overflows()` task-method for use with new
  `._allow_overruns: bool = True` mode.
 - add back a `._scope_nursery: trio.Nursery` (allocated in
   `Portal.open_context()`) who's sole purpose is to spawn a single task
   which runs the above method; anything else is an error.
 - augment `._deliver_msg()` to start a task and run the above method
   when operating in no overrun mode; the task queues overflow msgs and
   attempts to send them to the underlying mem chan using a blocking
   `.send()` call.
 - on context exit, any existing "drainer task" will be cancelled and
   remaining overflow queued msgs are discarded with a warning.
- rename `._error` -> `_remote_error` and set it in a new method
  `_maybe_cancel_and_set_remote_error()` which is called before
  processing
- adjust `.result()` to always call `._maybe_raise_remote_err()` at its
  start such that whenever a `ContextCancelled` arrives we do logic for
  whether or not to immediately raise that error or ignore it due to the
  current actor being the one who requested the cancel, by checking the
  error's `.canceller` field.
 - set the default value of `._result` to be `id(Context()` thus avoiding
   conflict with any `.result()` actually being `False`..

._runtime.Actor:
- augment `.cancel()` and `._cancel_task()` and `.cancel_rpc_tasks()` to
  take a `requesting_uid: tuple` indicating the source actor of every
  cancellation request.
- pass through the new `Context._allow_overruns` through `.get_context()`
- call the new `Context._deliver_msg()` from `._push_result()` (since
  the factoring out that method's contents).

._runtime._invoke:
- `TastStatus.started()` back a `Context` (unless an error is raised)
  instead of the cancel scope to make it easy to set/get state on that
  context for the purposes of cancellation and remote error relay.
- always raise any remote error via `Context._maybe_raise_remote_err()`
  before doing any `ContextCancelled` logic.
- assign any `Context._cancel_called_remote` set by the `requesting_uid`
  cancel methods (mentioned above) to the `ContextCancelled.canceller`.

._runtime.process_messages:
- always pass a `requesting_uid: tuple` to `Actor.cancel()` and
  `._cancel_task` to that any corresponding `ContextCancelled.canceller`
  can be set inside `._invoke()`.
ctx_cancel_semantics_and_overruns
Tyler Goodlet 2023-04-13 16:03:35 -04:00
parent 90e41016b9
commit c72026091e
4 changed files with 724 additions and 303 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

88
tractor/_portal.py
View File

@ -103,7 +103,7 @@ class Portal:
# When set to a ``Context`` (when _submit_for_result is called)
# it is expected that ``result()`` will be awaited at some
# point.
self._expect_result: Optional[Context] = None
self._expect_result: Context | None = None
self._streams: set[MsgStream] = set()
self.actor = current_actor()
@ -209,7 +209,10 @@ class Portal:
try:
# send cancel cmd - might not get response
# XXX: sure would be nice to make this work with a proper shield
with trio.move_on_after(timeout or self.cancel_timeout) as cs:
with trio.move_on_after(
timeout
or self.cancel_timeout
) as cs:
cs.shield = True
await self.run_from_ns('self', 'cancel')
@ -330,7 +333,9 @@ class Portal:
f'{async_gen_func} must be an async generator function!')
fn_mod_path, fn_name = NamespacePath.from_ref(
async_gen_func).to_tuple()
async_gen_func
).to_tuple()
ctx = await self.actor.start_remote_task(
self.channel,
fn_mod_path,
@ -396,13 +401,16 @@ class Portal:
raise TypeError(
f'{func} must be an async generator function!')
# TODO: i think from here onward should probably
# just be factored into an `@acm` inside a new
# a new `_context.py` mod.
fn_mod_path, fn_name = NamespacePath.from_ref(func).to_tuple()
ctx = await self.actor.start_remote_task(
self.channel,
fn_mod_path,
fn_name,
kwargs
kwargs,
)
assert ctx._remote_func_type == 'context'
@ -426,29 +434,47 @@ class Portal:
f' but received a non-error msg:\n{pformat(msg)}'
)
_err: Optional[BaseException] = None
ctx._portal = self
_err: BaseException | None = None
ctx._portal: Portal = self
uid = self.channel.uid
cid = ctx.cid
etype: Optional[Type[BaseException]] = None
uid: tuple = self.channel.uid
cid: str = ctx.cid
etype: Type[BaseException] | None = None
# deliver context instance and .started() msg value in open tuple.
# deliver context instance and .started() msg value in enter
# tuple.
try:
async with trio.open_nursery() as scope_nursery:
ctx._scope_nursery = scope_nursery
# do we need this?
# await trio.lowlevel.checkpoint()
async with trio.open_nursery() as nurse:
ctx._scope_nursery = nurse
ctx._scope = nurse.cancel_scope
yield ctx, first
# when in allow_ovveruns mode there may be lingering
# overflow sender tasks remaining?
if nurse.child_tasks:
# ensure we are in overrun state with
# ``._allow_overruns=True`` bc otherwise
# there should be no tasks in this nursery!
if (
not ctx._allow_overruns
or len(nurse.child_tasks) > 1
):
raise RuntimeError(
'Context has sub-tasks but is '
'not in `allow_overruns=True` Mode!?'
)
ctx._scope.cancel()
except ContextCancelled as err:
_err = err
# swallow and mask cross-actor task context cancels that
# were initiated by *this* side's task.
if not ctx._cancel_called:
# context was cancelled at the far end but was
# not part of this end requesting that cancel
# so raise for the local task to respond and handle.
# XXX: this should NEVER happen!
# from ._debug import breakpoint
# await breakpoint()
raise
# if the context was cancelled by client code
@ -468,17 +494,17 @@ class Portal:
) as err:
etype = type(err)
# the context cancels itself on any cancel
# causing error.
if ctx.chan.connected():
log.cancel(
'Context cancelled for task, sending cancel request..\n'
f'task:{cid}\n'
f'actor:{uid}'
)
# cancel ourselves on any error.
log.cancel(
'Context cancelled for task, sending cancel request..\n'
f'task:{cid}\n'
f'actor:{uid}'
)
try:
await ctx.cancel()
else:
except trio.BrokenResourceError:
log.warning(
'IPC connection for context is broken?\n'
f'task:{cid}\n'
@ -487,12 +513,7 @@ class Portal:
raise
finally:
# in the case where a runtime nursery (due to internal bug)
# or a remote actor transmits an error we want to be
# sure we get the error the underlying feeder mem chan.
# if it's not raised here it *should* be raised from the
# msg loop nursery right?
else:
if ctx.chan.connected():
log.info(
'Waiting on final context-task result for\n'
@ -505,6 +526,7 @@ class Portal:
f'value from callee `{result}`'
)
finally:
# though it should be impossible for any tasks
# operating *in* this scope to have survived
# we tear down the runtime feeder chan last

4
tractor/_root.py
View File

@ -254,7 +254,9 @@ async def open_root_actor(
# tempn.start_soon(an.exited.wait)
logger.cancel("Shutting down root actor")
await actor.cancel()
await actor.cancel(
requesting_uid=actor.uid,
)
finally:
_state._current_actor = None

417
tractor/_runtime.py
View File

@ -28,9 +28,11 @@ import inspect
import signal
import sys
from typing import (
Any, Optional,
Union, TYPE_CHECKING,
Any,
Callable,
Optional,
Union,
TYPE_CHECKING,
)
import uuid
from types import ModuleType
@ -44,7 +46,10 @@ import trio # type: ignore
from trio_typing import TaskStatus
from ._ipc import Channel
from ._streaming import Context
from ._streaming import (
mk_context,
Context,
)
from .log import get_logger
from ._exceptions import (
pack_error,
@ -53,7 +58,6 @@ from ._exceptions import (
is_multi_cancelled,
ContextCancelled,
TransportClosed,
StreamOverrun,
)
from . import _debug
from ._discovery import get_arbiter
@ -79,7 +83,7 @@ async def _invoke(
is_rpc: bool = True,
task_status: TaskStatus[
Union[trio.CancelScope, BaseException]
Union[Context, BaseException]
] = trio.TASK_STATUS_IGNORED,
):
'''
@ -99,7 +103,14 @@ async def _invoke(
# activated cancel scope ref
cs: Optional[trio.CancelScope] = None
ctx = actor.get_context(chan, cid)
ctx = actor.get_context(
chan,
cid,
# We shouldn't ever need to pass this through right?
# it's up to the soon-to-be called rpc task to
# open the stream with this option.
# allow_overruns=True,
)
context: bool = False
if getattr(func, '_tractor_stream_function', False):
@ -138,7 +149,10 @@ async def _invoke(
):
raise TypeError(f'{func} must be an async function!')
coro = func(**kwargs)
try:
coro = func(**kwargs)
except TypeError:
raise
if inspect.isasyncgen(coro):
await chan.send({'functype': 'asyncgen', 'cid': cid})
@ -150,7 +164,8 @@ async def _invoke(
# of the async gen in order to be sure the cancel
# is propagated!
with cancel_scope as cs:
task_status.started(cs)
ctx._scope = cs
task_status.started(ctx)
async with aclosing(coro) as agen:
async for item in agen:
# TODO: can we send values back in here?
@ -176,7 +191,8 @@ async def _invoke(
# manualy construct the response dict-packet-responses as
# above
with cancel_scope as cs:
task_status.started(cs)
ctx._scope = cs
task_status.started(ctx)
await coro
if not cs.cancelled_caught:
@ -189,19 +205,25 @@ async def _invoke(
await chan.send({'functype': 'context', 'cid': cid})
try:
async with trio.open_nursery() as scope_nursery:
ctx._scope_nursery = scope_nursery
cs = scope_nursery.cancel_scope
task_status.started(cs)
with cancel_scope as cs:
ctx._scope = cs
task_status.started(ctx)
res = await coro
await chan.send({'return': res, 'cid': cid})
except BaseExceptionGroup:
# XXX: do we ever trigger this block any more?
except (
BaseExceptionGroup,
trio.Cancelled,
):
# if a context error was set then likely
# thei multierror was raised due to that
if ctx._error is not None:
raise ctx._error from None
if ctx._remote_error is not None:
raise ctx._remote_error
# maybe TODO: pack in ``trio.Cancelled.__traceback__`` here
# so they can be unwrapped and displayed on the caller
# side?
raise
finally:
@ -213,7 +235,6 @@ async def _invoke(
# associated child isn't in debug any more
await _debug.maybe_wait_for_debugger()
ctx = actor._contexts.pop((chan.uid, cid))
if ctx:
log.runtime(
f'Context entrypoint {func} was terminated:\n{ctx}'
@ -221,32 +242,60 @@ async def _invoke(
assert cs
if cs.cancelled_caught:
# if 'brokerd.kraken' in actor.uid:
# await _debug.breakpoint()
# TODO: pack in ``trio.Cancelled.__traceback__`` here
# so they can be unwrapped and displayed on the caller
# side!
# first check for and raise any remote error
# before raising any context cancelled case
# so that real remote errors don't get masked as
# ``ContextCancelled``s.
re = ctx._remote_error
if re:
ctx._maybe_raise_remote_err(re)
fname = func.__name__
if ctx._cancel_called:
msg = f'`{fname}()`@{actor.uid} cancelled itself'
if cs.cancel_called:
canceller = ctx._cancel_called_remote
# await _debug.breakpoint()
elif cs.cancel_called:
msg = (
f'`{fname}()`@{actor.uid} was remotely cancelled by its caller '
f'{ctx.chan.uid}'
# NOTE / TODO: if we end up having
# ``Actor._cancel_task()`` call
# ``Context.cancel()`` directly, we're going to
# need to change this logic branch since it will
# always enter..
if ctx._cancel_called:
msg = f'`{fname}()`@{actor.uid} cancelled itself'
else:
msg = (
f'`{fname}()`@{actor.uid} '
'was remotely cancelled by '
)
# if the channel which spawned the ctx is the
# one that cancelled it then we report that, vs.
# it being some other random actor that for ex.
# some actor who calls `Portal.cancel_actor()`
# and by side-effect cancels this ctx.
if canceller == ctx.chan.uid:
msg += f'its caller {canceller}'
else:
msg += f'remote actor {canceller}'
# TODO: does this ever get set any more or can
# we remove it?
if ctx._cancel_msg:
msg += f' with msg:\n{ctx._cancel_msg}'
# task-contex was either cancelled by request using
# ``Portal.cancel_actor()`` or ``Context.cancel()``
# on the far end, or it was cancelled by the local
# (callee) task, so relay this cancel signal to the
# other side.
raise ContextCancelled(
msg,
suberror_type=trio.Cancelled,
canceller=canceller,
)
if ctx._cancel_msg:
msg += f' with msg:\n{ctx._cancel_msg}'
# task-contex was cancelled so relay to the cancel to caller
raise ContextCancelled(
msg,
suberror_type=trio.Cancelled,
)
else:
# regular async function
try:
@ -261,12 +310,17 @@ async def _invoke(
)
with cancel_scope as cs:
task_status.started(cs)
ctx._scope = cs
task_status.started(ctx)
result = await coro
log.cancel(f'result: {result}')
fname = func.__name__
log.runtime(f'{fname}() result: {result}')
if not failed_resp:
# only send result if we know IPC isn't down
await chan.send({'return': result, 'cid': cid})
await chan.send(
{'return': result,
'cid': cid}
)
except (
Exception,
@ -309,6 +363,7 @@ async def _invoke(
# always ship errors back to caller
err_msg = pack_error(err, tb=tb)
err_msg['cid'] = cid
try:
await chan.send(err_msg)
@ -325,14 +380,21 @@ async def _invoke(
f"Failed to ship error to caller @ {chan.uid} !?"
)
# error is probably from above coro running code *not from the
# underlyingn rpc invocation* since a scope was never allocated
# around actual coroutine await.
if cs is None:
# error is from above code not from rpc invocation
# we don't ever raise directly here to allow the
# msg-loop-scheduler to continue running for this
# channel.
task_status.started(err)
finally:
# RPC task bookeeping
try:
scope, func, is_complete = actor._rpc_tasks.pop((chan, cid))
ctx, func, is_complete = actor._rpc_tasks.pop(
(chan, cid)
)
is_complete.set()
except KeyError:
@ -341,6 +403,9 @@ async def _invoke(
# cancel scope will not have been inserted yet
log.warning(
f"Task {func} likely errored or cancelled before start")
else:
log.cancel(f'{func.__name__}({kwargs}) failed?')
finally:
if not actor._rpc_tasks:
log.runtime("All RPC tasks have completed")
@ -439,6 +504,7 @@ class Actor:
self.uid = (name, uid or str(uuid.uuid4()))
self._cancel_complete = trio.Event()
self._cancel_called_remote: tuple[str, tuple] | None = None
self._cancel_called: bool = False
# retreive and store parent `__main__` data which
@ -477,7 +543,7 @@ class Actor:
# (chan, cid) -> (cancel_scope, func)
self._rpc_tasks: dict[
tuple[Channel, str],
tuple[trio.CancelScope, Callable, trio.Event]
tuple[Context, Callable, trio.Event]
] = {}
# map {actor uids -> Context}
@ -652,8 +718,8 @@ class Actor:
if (
local_nursery
):
log.cancel(f"Waiting on cancel request to peer {chan.uid}")
if chan._cancel_called:
log.cancel(f"Waiting on cancel request to peer {chan.uid}")
# XXX: this is a soft wait on the channel (and its
# underlying transport protocol) to close from the
# remote peer side since we presume that any channel
@ -786,76 +852,15 @@ class Actor:
f'\n{msg}')
return
send_chan = ctx._send_chan
log.runtime(f"Delivering {msg} from {chan.uid} to caller {cid}")
# XXX: we do **not** maintain backpressure and instead
# opt to relay stream overrun errors to the sender.
try:
send_chan.send_nowait(msg)
# if an error is deteced we should always
# expect it to be raised by any context (stream)
# consumer task
await ctx._maybe_raise_from_remote_msg(msg)
except trio.BrokenResourceError:
# TODO: what is the right way to handle the case where the
# local task has already sent a 'stop' / StopAsyncInteration
# to the other side but and possibly has closed the local
# feeder mem chan? Do we wait for some kind of ack or just
# let this fail silently and bubble up (currently)?
# XXX: local consumer has closed their side
# so cancel the far end streaming task
log.warning(f"{send_chan} consumer is already closed")
return
except trio.WouldBlock:
# XXX: always push an error even if the local
# receiver is in overrun state.
await ctx._maybe_raise_from_remote_msg(msg)
uid = chan.uid
lines = [
'Task context stream was overrun',
f'local task: {cid} @ {self.uid}',
f'remote sender: {uid}',
]
if not ctx._stream_opened:
lines.insert(
1,
f'\n*** No stream open on `{self.uid[0]}` side! ***\n'
)
text = '\n'.join(lines)
if ctx._backpressure:
log.warning(text)
try:
await send_chan.send(msg)
except trio.BrokenResourceError:
# XXX: local consumer has closed their side
# so cancel the far end streaming task
log.warning(f"{chan} is already closed")
else:
try:
raise StreamOverrun(text) from None
except StreamOverrun as err:
err_msg = pack_error(err)
err_msg['cid'] = cid
try:
await chan.send(err_msg)
except trio.BrokenResourceError:
# XXX: local consumer has closed their side
# so cancel the far end streaming task
log.warning(f"{chan} is already closed")
return await ctx._deliver_msg(msg)
def get_context(
self,
chan: Channel,
cid: str,
msg_buffer_size: Optional[int] = None,
msg_buffer_size: int | None = None,
allow_overruns: bool = False,
) -> Context:
'''
@ -871,6 +876,7 @@ class Actor:
assert actor_uid
try:
ctx = self._contexts[(actor_uid, cid)]
ctx._allow_overruns = allow_overruns
# adjust buffer size if specified
state = ctx._send_chan._state # type: ignore
@ -878,15 +884,11 @@ class Actor:
state.max_buffer_size = msg_buffer_size
except KeyError:
send_chan: trio.MemorySendChannel
recv_chan: trio.MemoryReceiveChannel
send_chan, recv_chan = trio.open_memory_channel(
msg_buffer_size or self.msg_buffer_size)
ctx = Context(
ctx = mk_context(
chan,
cid,
_send_chan=send_chan,
_recv_chan=recv_chan,
msg_buffer_size=msg_buffer_size or self.msg_buffer_size,
_allow_overruns=allow_overruns,
)
self._contexts[(actor_uid, cid)] = ctx
@ -898,7 +900,8 @@ class Actor:
ns: str,
func: str,
kwargs: dict,
msg_buffer_size: Optional[int] = None,
msg_buffer_size: int | None = None,
allow_overruns: bool = False,
) -> Context:
'''
@ -916,6 +919,7 @@ class Actor:
chan,
cid,
msg_buffer_size=msg_buffer_size,
allow_overruns=allow_overruns,
)
log.runtime(f"Sending cmd to {chan.uid}: {ns}.{func}({kwargs})")
await chan.send(
@ -1046,7 +1050,11 @@ class Actor:
assert self._service_n
self._service_n.start_soon(self.cancel)
async def cancel(self) -> bool:
async def cancel(
self,
requesting_uid: tuple[str, str],
) -> bool:
'''
Cancel this actor's runtime.
@ -1060,6 +1068,7 @@ class Actor:
'''
log.cancel(f"{self.uid} is trying to cancel")
self._cancel_called_remote: tuple = requesting_uid
self._cancel_called = True
# cancel all ongoing rpc tasks
@ -1073,7 +1082,7 @@ class Actor:
dbcs.cancel()
# kill all ongoing tasks
await self.cancel_rpc_tasks()
await self.cancel_rpc_tasks(requesting_uid=requesting_uid)
# stop channel server
self.cancel_server()
@ -1099,7 +1108,13 @@ class Actor:
# for n in root.child_nurseries:
# n.cancel_scope.cancel()
async def _cancel_task(self, cid, chan):
async def _cancel_task(
self,
cid: str,
chan: Channel,
requesting_uid: tuple[str, str] | None = None,
) -> bool:
'''
Cancel a local task by call-id / channel.
@ -1114,35 +1129,51 @@ class Actor:
try:
# this ctx based lookup ensures the requested task to
# be cancelled was indeed spawned by a request from this channel
scope, func, is_complete = self._rpc_tasks[(chan, cid)]
ctx, func, is_complete = self._rpc_tasks[(chan, cid)]
scope = ctx._scope
except KeyError:
log.cancel(f"{cid} has already completed/terminated?")
return
return True
log.cancel(
f"Cancelling task:\ncid: {cid}\nfunc: {func}\n"
f"peer: {chan.uid}\n")
if (
ctx._cancel_called_remote is None
and requesting_uid
):
ctx._cancel_called_remote: tuple = requesting_uid
# don't allow cancelling this function mid-execution
# (is this necessary?)
if func is self._cancel_task:
return
return True
# TODO: shouldn't we eventually be calling ``Context.cancel()``
# directly here instead (since that method can handle both
# side's calls into it?
scope.cancel()
# wait for _invoke to mark the task complete
log.runtime(
f"Waiting on task to cancel:\ncid: {cid}\nfunc: {func}\n"
f"peer: {chan.uid}\n")
'Waiting on task to cancel:\n'
f'cid: {cid}\nfunc: {func}\n'
f'peer: {chan.uid}\n'
)
await is_complete.wait()
log.runtime(
f"Sucessfully cancelled task:\ncid: {cid}\nfunc: {func}\n"
f"peer: {chan.uid}\n")
return True
async def cancel_rpc_tasks(
self,
only_chan: Channel | None = None,
requesting_uid: tuple[str, str] | None = None,
) -> None:
'''
Cancel all existing RPC responder tasks using the cancel scope
@ -1154,7 +1185,7 @@ class Actor:
log.cancel(f"Cancelling all {len(tasks)} rpc tasks:\n{tasks} ")
for (
(chan, cid),
(scope, func, is_complete),
(ctx, func, is_complete),
) in tasks.copy().items():
if only_chan is not None:
if only_chan != chan:
@ -1162,7 +1193,11 @@ class Actor:
# TODO: this should really done in a nursery batch
if func != self._cancel_task:
await self._cancel_task(cid, chan)
await self._cancel_task(
cid,
chan,
requesting_uid=requesting_uid,
)
log.cancel(
f"Waiting for remaining rpc tasks to complete {tasks}")
@ -1248,8 +1283,8 @@ async def async_main(
Actor runtime entrypoint; start the IPC channel server, maybe connect
back to the parent, and startup all core machinery tasks.
A "root-most" (or "top-level") nursery for this actor is opened here
and when cancelled effectively cancels the actor.
A "root" (or "top-level") nursery for this actor is opened here and
when cancelled/terminated effectively closes the actor's "runtime".
'''
# attempt to retreive ``trio``'s sigint handler and stash it
@ -1446,15 +1481,16 @@ async def process_messages(
) -> bool:
'''
Process messages for the IPC transport channel async-RPC style.
This is the per-channel, low level RPC task scheduler loop.
Receive multiplexed RPC requests, spawn handler tasks and deliver
responses over or boxed errors back to the "caller" task.
Receive multiplexed RPC request messages from some remote process,
spawn handler tasks depending on request type and deliver responses
or boxed errors back to the remote caller (task).
'''
# TODO: once https://github.com/python-trio/trio/issues/467 gets
# worked out we'll likely want to use that!
msg = None
msg: dict | None = None
nursery_cancelled_before_task: bool = False
log.runtime(f"Entering msg loop for {chan} from {chan.uid}")
@ -1476,7 +1512,10 @@ async def process_messages(
for (channel, cid) in actor._rpc_tasks.copy():
if channel is chan:
await actor._cancel_task(cid, channel)
await actor._cancel_task(
cid,
channel,
)
log.runtime(
f"Msg loop signalled to terminate for"
@ -1490,12 +1529,14 @@ async def process_messages(
cid = msg.get('cid')
if cid:
# deliver response to local caller/waiter
# via its per-remote-context memory channel.
await actor._push_result(chan, cid, msg)
log.runtime(
f"Waiting on next msg for {chan} from {chan.uid}")
continue
# TODO: implement with ``match:`` syntax?
# process command request
try:
ns, funcname, kwargs, actorid, cid = msg['cmd']
@ -1515,13 +1556,12 @@ async def process_messages(
f"{ns}.{funcname}({kwargs})")
if ns == 'self':
func = getattr(actor, funcname)
if funcname == 'cancel':
func = actor.cancel
kwargs['requesting_uid'] = chan.uid
# don't start entire actor runtime
# cancellation if this actor is in debug
# mode
# don't start entire actor runtime cancellation
# if this actor is currently in debug mode!
pdb_complete = _debug.Lock.local_pdb_complete
if pdb_complete:
await pdb_complete.wait()
@ -1533,43 +1573,56 @@ async def process_messages(
# msg loop and break out into
# ``async_main()``
log.cancel(
f"Actor {actor.uid} was remotely cancelled "
"Actor runtime for was remotely cancelled "
f"by {chan.uid}"
)
await _invoke(
actor, cid, chan, func, kwargs, is_rpc=False
actor,
cid,
chan,
func,
kwargs,
is_rpc=False,
)
log.cancel(
f'Cancelling msg loop for {chan.uid}'
)
loop_cs.cancel()
break
if funcname == '_cancel_task':
func = actor._cancel_task
# we immediately start the runtime machinery
# shutdown
with trio.CancelScope(shield=True):
# actor.cancel() was called so kill this
# msg loop and break out into
# ``async_main()``
kwargs['chan'] = chan
log.cancel(
f'Remote request to cancel task\n'
f'remote actor: {chan.uid}\n'
f'task: {cid}'
# with trio.CancelScope(shield=True):
kwargs['chan'] = chan
target_cid = kwargs['cid']
kwargs['requesting_uid'] = chan.uid
log.cancel(
f'Remote request to cancel task\n'
f'remote actor: {chan.uid}\n'
f'task: {target_cid}'
)
try:
await _invoke(
actor,
cid,
chan,
func,
kwargs,
is_rpc=False,
)
try:
await _invoke(
actor,
cid,
chan,
func,
kwargs,
is_rpc=False,
)
except BaseException:
log.exception("failed to cancel task?")
except BaseException:
log.exception("failed to cancel task?")
continue
else:
# normally registry methods, eg.
# ``.register_actor()`` etc.
func = getattr(actor, funcname)
continue
else:
# complain to client about restricted modules
try:
@ -1584,34 +1637,49 @@ async def process_messages(
log.runtime(f"Spawning task for {func}")
assert actor._service_n
try:
cs = await actor._service_n.start(
partial(_invoke, actor, cid, chan, func, kwargs),
ctx: Context = await actor._service_n.start(
partial(
_invoke,
actor,
cid,
chan,
func,
kwargs,
),
name=funcname,
)
except (
RuntimeError,
BaseExceptionGroup,
):
# avoid reporting a benign race condition
# during actor runtime teardown.
nursery_cancelled_before_task = True
nursery_cancelled_before_task: bool = True
break
# never allow cancelling cancel requests (results in
# deadlock and other weird behaviour)
# if func != actor.cancel:
if isinstance(cs, Exception):
# in the lone case where a ``Context`` is not
# delivered, it's likely going to be a locally
# scoped exception from ``_invoke()`` itself.
if isinstance(ctx, Exception):
log.warning(
f"Task for RPC func {func} failed with"
f"{cs}")
f"{ctx}"
)
continue
else:
# mark that we have ongoing rpc tasks
actor._ongoing_rpc_tasks = trio.Event()
log.runtime(f"RPC func is {func}")
# store cancel scope such that the rpc task can be
# cancelled gracefully if requested
actor._rpc_tasks[(chan, cid)] = (
cs, func, trio.Event())
ctx,
func,
trio.Event(),
)
log.runtime(
f"Waiting on next msg for {chan} from {chan.uid}")
@ -1655,7 +1723,7 @@ async def process_messages(
match err:
case ContextCancelled():
log.cancel(
f'Actor: {actor.uid} was task-context-cancelled with,\n'
f'Actor: {actor.uid} was context-cancelled with,\n'
f'str(err)'
)
case _:
@ -1672,7 +1740,8 @@ async def process_messages(
# msg debugging for when he machinery is brokey
log.runtime(
f"Exiting msg loop for {chan} from {chan.uid} "
f"with last msg:\n{msg}")
f"with last msg:\n{msg}"
)
# transport **was not** disconnected
return False

518
tractor/_streaming.py
View File

@ -21,24 +21,41 @@ Message stream types and APIs.
from __future__ import annotations
import inspect
from contextlib import asynccontextmanager
from dataclasses import dataclass
from collections import deque
from dataclasses import (
dataclass,
field,
)
from functools import partial
from pprint import pformat
from typing import (
Any,
Optional,
Callable,
AsyncGenerator,
AsyncIterator
AsyncIterator,
TYPE_CHECKING,
)
import warnings
import trio
from ._ipc import Channel
from ._exceptions import unpack_error, ContextCancelled
from ._state import current_actor
from ._exceptions import (
unpack_error,
pack_error,
ContextCancelled,
StreamOverrun,
)
from .log import get_logger
from .trionics import broadcast_receiver, BroadcastReceiver
from ._state import current_actor
from .trionics import (
broadcast_receiver,
BroadcastReceiver,
)
if TYPE_CHECKING:
from ._portal import Portal
log = get_logger(__name__)
@ -335,8 +352,8 @@ class MsgStream(trio.abc.Channel):
Send a message over this stream to the far end.
'''
if self._ctx._error:
raise self._ctx._error # from None
if self._ctx._remote_error:
raise self._ctx._remote_error # from None
if self._closed:
raise trio.ClosedResourceError('This stream was already closed')
@ -373,25 +390,61 @@ class Context:
_recv_chan: trio.MemoryReceiveChannel
_send_chan: trio.MemorySendChannel
_remote_func_type: Optional[str] = None
_remote_func_type: str | None = None
# only set on the caller side
_portal: Optional['Portal'] = None # type: ignore # noqa
_result: Optional[Any] = False
_error: Optional[BaseException] = None
_portal: Portal | None = None # type: ignore # noqa
_result: Any | int = None
_remote_error: BaseException | None = None
# status flags
# cancellation state
_cancel_called: bool = False
_cancel_msg: Optional[str] = None
_cancel_called_remote: tuple | None = None
_cancel_msg: str | None = None
_scope: trio.CancelScope | None = None
_enter_debugger_on_cancel: bool = True
@property
def cancel_called(self) -> bool:
'''
Records whether cancellation has been requested for this context
by either an explicit call to ``.cancel()`` or an implicit call
due to an error caught inside the ``Portal.open_context()``
block.
'''
return self._cancel_called
@property
def cancel_called_remote(self) -> tuple[str, str] | None:
'''
``Actor.uid`` of the remote actor who's task was cancelled
causing this side of the context to also be cancelled.
'''
remote_uid = self._cancel_called_remote
if remote_uid:
return tuple(remote_uid)
# init and streaming state
_started_called: bool = False
_started_received: bool = False
_stream_opened: bool = False
# only set on the callee side
_scope_nursery: Optional[trio.Nursery] = None
_backpressure: bool = True
# overrun handling machinery
# NOTE: none of this provides "backpressure" to the remote
# task, only an ability to not lose messages when the local
# task is configured to NOT transmit ``StreamOverrun``s back
# to the other side.
_overflow_q: deque[dict] = field(
default_factory=partial(
deque,
maxlen=616,
)
)
_scope_nursery: trio.Nursery | None = None
_in_overrun: bool = False
_allow_overruns: bool = False
async def send_yield(
self,
@ -410,9 +463,9 @@ class Context:
async def send_stop(self) -> None:
await self.chan.send({'stop': True, 'cid': self.cid})
async def _maybe_raise_from_remote_msg(
async def _maybe_cancel_and_set_remote_error(
self,
msg: dict[str, Any],
error_msg: dict[str, Any],
) -> None:
'''
@ -423,55 +476,77 @@ class Context:
in the corresponding remote callee task.
'''
error = msg.get('error')
if error:
# If this is an error message from a context opened by
# ``Portal.open_context()`` we want to interrupt any ongoing
# (child) tasks within that context to be notified of the remote
# error relayed here.
#
# The reason we may want to raise the remote error immediately
# is that there is no guarantee the associated local task(s)
# will attempt to read from any locally opened stream any time
# soon.
#
# NOTE: this only applies when
# ``Portal.open_context()`` has been called since it is assumed
# (currently) that other portal APIs (``Portal.run()``,
# ``.run_in_actor()``) do their own error checking at the point
# of the call and result processing.
log.error(
f'Remote context error for {self.chan.uid}:{self.cid}:\n'
f'{msg["error"]["tb_str"]}'
# If this is an error message from a context opened by
# ``Portal.open_context()`` we want to interrupt any ongoing
# (child) tasks within that context to be notified of the remote
# error relayed here.
#
# The reason we may want to raise the remote error immediately
# is that there is no guarantee the associated local task(s)
# will attempt to read from any locally opened stream any time
# soon.
#
# NOTE: this only applies when
# ``Portal.open_context()`` has been called since it is assumed
# (currently) that other portal APIs (``Portal.run()``,
# ``.run_in_actor()``) do their own error checking at the point
# of the call and result processing.
error = unpack_error(
error_msg,
self.chan,
)
# XXX: set the remote side's error so that after we cancel
# whatever task is the opener of this context it can raise
# that error as the reason.
self._remote_error = error
if (
isinstance(error, ContextCancelled)
):
log.cancel(
'Remote task-context sucessfully cancelled for '
f'{self.chan.uid}:{self.cid}'
)
error = unpack_error(msg, self.chan)
if (
isinstance(error, ContextCancelled) and
self._cancel_called
):
if self._cancel_called:
# this is an expected cancel request response message
# and we don't need to raise it in scope since it will
# potentially override a real error
return
else:
log.error(
f'Remote context error for {self.chan.uid}:{self.cid}:\n'
f'{error_msg["error"]["tb_str"]}'
)
# TODO: tempted to **not** do this by-reraising in a
# nursery and instead cancel a surrounding scope, detect
# the cancellation, then lookup the error that was set?
# YES! this is way better and simpler!
if (
self._scope
):
# from trio.testing import wait_all_tasks_blocked
# await wait_all_tasks_blocked()
self._cancel_called_remote = self.chan.uid
self._scope.cancel()
self._error = error
# NOTE: this usage actually works here B)
# from ._debug import breakpoint
# await breakpoint()
# TODO: tempted to **not** do this by-reraising in a
# nursery and instead cancel a surrounding scope, detect
# the cancellation, then lookup the error that was set?
if self._scope_nursery:
async def raiser():
raise self._error from None
# from trio.testing import wait_all_tasks_blocked
# await wait_all_tasks_blocked()
if not self._scope_nursery._closed: # type: ignore
self._scope_nursery.start_soon(raiser)
# XXX: this will break early callee results sending
# since when `.result()` is finally called, this
# chan will be closed..
# if self._recv_chan:
# await self._recv_chan.aclose()
async def cancel(
self,
msg: str | None = None,
timeout: float = 0.5,
# timeout: float = 1000,
) -> None:
'''
@ -488,6 +563,8 @@ class Context:
log.cancel(f'Cancelling {side} side of context to {self.chan.uid}')
self._cancel_called = True
# await _debug.breakpoint()
# breakpoint()
if side == 'caller':
if not self._portal:
@ -496,8 +573,8 @@ class Context:
)
cid = self.cid
with trio.move_on_after(0.5) as cs:
cs.shield = True
with trio.move_on_after(timeout) as cs:
# cs.shield = True
log.cancel(
f"Cancelling stream {cid} to "
f"{self._portal.channel.uid}")
@ -505,7 +582,12 @@ class Context:
# NOTE: we're telling the far end actor to cancel a task
# corresponding to *this actor*. The far end local channel
# instance is passed to `Actor._cancel_task()` implicitly.
await self._portal.run_from_ns('self', '_cancel_task', cid=cid)
await self._portal.run_from_ns(
'self',
'_cancel_task',
cid=cid,
)
# print("EXITING CANCEL CALL")
if cs.cancelled_caught:
# XXX: there's no way to know if the remote task was indeed
@ -530,17 +612,14 @@ class Context:
# {'error': trio.Cancelled, cid: "blah"} enough?
# This probably gets into the discussion in
# https://github.com/goodboy/tractor/issues/36
assert self._scope_nursery
self._scope_nursery.cancel_scope.cancel()
if self._recv_chan:
await self._recv_chan.aclose()
assert self._scope
self._scope.cancel()
@asynccontextmanager
async def open_stream(
self,
backpressure: bool | None = True,
allow_overruns: bool | None = False,
msg_buffer_size: int | None = None,
) -> AsyncGenerator[MsgStream, None]:
@ -592,8 +671,9 @@ class Context:
self.chan,
self.cid,
msg_buffer_size=msg_buffer_size,
allow_overruns=allow_overruns,
)
ctx._backpressure = backpressure
ctx._allow_overruns = allow_overruns
assert ctx is self
# XXX: If the underlying channel feeder receive mem chan has
@ -637,48 +717,115 @@ class Context:
f'ctx id: {self.cid}'
)
async def result(self) -> Any:
def _maybe_raise_remote_err(
self,
err: Exception,
) -> None:
# NOTE: whenever the context's "opener" side (task) **is**
# the side which requested the cancellation (likekly via
# ``Context.cancel()``), we don't want to re-raise that
# cancellation signal locally (would be akin to
# a ``trio.Nursery`` nursery raising ``trio.Cancelled``
# whenever ``CancelScope.cancel()`` was called) and instead
# silently reap the expected cancellation "error"-msg.
# if 'pikerd' in err.msgdata['tb_str']:
# # from . import _debug
# # await _debug.breakpoint()
# breakpoint()
if (
isinstance(err, ContextCancelled)
and (
self._cancel_called
or self.chan._cancel_called
or tuple(err.canceller) == current_actor().uid
)
):
return err
raise err from None
async def result(self) -> Any | Exception:
'''
From a caller side, wait for and return the final result from
the callee side task.
From some (caller) side task, wait for and return the final
result from the remote (callee) side's task.
This provides a mechanism for one task running in some actor to wait
on another task at the other side, in some other actor, to terminate.
If the remote task is still in a streaming state (it is delivering
values from inside a ``Context.open_stream():`` block, then those
msgs are drained but discarded since it is presumed this side of
the context has already finished with its own streaming logic.
If the remote context (or its containing actor runtime) was
canceled, either by a local task calling one of
``Context.cancel()`` or `Portal.cancel_actor()``, we ignore the
received ``ContextCancelled`` exception if the context or
underlying IPC channel is marked as having been "cancel called".
This is similar behavior to using ``trio.Nursery.cancel()``
wherein tasks which raise ``trio.Cancel`` are silently reaped;
the main different in this API is in the "cancel called" case,
instead of just not raising, we also return the exception *as
the result* since client code may be interested in the details
of the remote cancellation.
'''
assert self._portal, "Context.result() can not be called from callee!"
assert self._recv_chan
if self._result is False:
# from . import _debug
# await _debug.breakpoint()
if not self._recv_chan._closed: # type: ignore
re = self._remote_error
if re:
self._maybe_raise_remote_err(re)
return re
# wait for a final context result consuming
# and discarding any bi dir stream msgs still
# in transit from the far end.
while True:
if (
self._result == id(self)
and not self._remote_error
and not self._recv_chan._closed # type: ignore
):
# wait for a final context result consuming
# and discarding any bi dir stream msgs still
# in transit from the far end.
while True:
msg = await self._recv_chan.receive()
try:
self._result = msg['return']
msg = await self._recv_chan.receive()
try:
self._result = msg['return']
break
except KeyError as msgerr:
# NOTE: we don't need to do this right?
# XXX: only close the rx mem chan AFTER
# a final result is retreived.
# if self._recv_chan:
# await self._recv_chan.aclose()
if 'yield' in msg:
# far end task is still streaming to us so discard
log.warning(f'Discarding stream delivered {msg}')
continue
break
except KeyError: # as msgerr:
elif 'stop' in msg:
log.debug('Remote stream terminated')
continue
if 'yield' in msg:
# far end task is still streaming to us so discard
log.warning(f'Discarding stream delivered {msg}')
continue
# internal error should never get here
assert msg.get('cid'), (
"Received internal error at portal?")
elif 'stop' in msg:
log.debug('Remote stream terminated')
continue
raise unpack_error(
msg, self._portal.channel
) from msgerr
# internal error should never get here
assert msg.get('cid'), (
"Received internal error at portal?")
return self._result
err = unpack_error(
msg,
self._portal.channel
) # from msgerr
err = self._maybe_raise_remote_err(err)
self._remote_err = err
return self._remote_error or self._result
async def started(
self,
@ -708,6 +855,187 @@ class Context:
# async def restart(self) -> None:
# pass
async def _drain_overflows(
self,
) -> None:
'''
Private task spawned to push newly received msgs to the local
task which getting overrun by the remote side.
In order to not block the rpc msg loop, but also not discard
msgs received in this context, we need to async push msgs in
a new task which only runs for as long as the local task is in
an overrun state.
'''
self._in_overrun = True
try:
while self._overflow_q:
# NOTE: these msgs should never be errors since we always do
# the check prior to checking if we're in an overrun state
# inside ``.deliver_msg()``.
msg = self._overflow_q.popleft()
try:
await self._send_chan.send(msg)
except trio.BrokenResourceError:
log.warning(
f"{self._send_chan} consumer is already closed"
)
return
except trio.Cancelled:
# we are obviously still in overrun
# but the context is being closed anyway
# so we just warn that there are un received
# msgs still..
self._overflow_q.appendleft(msg)
fmt_msgs = ''
for msg in self._overflow_q:
fmt_msgs += f'{pformat(msg)}\n'
log.warning(
f'Context for {self.cid} is being closed while '
'in an overrun state!\n'
'Discarding the following msgs:\n'
f'{fmt_msgs}\n'
)
raise
finally:
# task is now finished with the backlog so mark us as
# no longer in backlog.
self._in_overrun = False
async def _deliver_msg(
self,
msg: dict,
draining: bool = False,
) -> bool:
cid = self.cid
chan = self.chan
uid = chan.uid
send_chan: trio.MemorySendChannel = self._send_chan
log.runtime(
f"Delivering {msg} from {uid} to caller {cid}"
)
error = msg.get('error')
if error:
await self._maybe_cancel_and_set_remote_error(msg)
if (
self._in_overrun
):
self._overflow_q.append(msg)
return False
try:
send_chan.send_nowait(msg)
return True
# if an error is deteced we should always
# expect it to be raised by any context (stream)
# consumer task
except trio.BrokenResourceError:
# TODO: what is the right way to handle the case where the
# local task has already sent a 'stop' / StopAsyncInteration
# to the other side but and possibly has closed the local
# feeder mem chan? Do we wait for some kind of ack or just
# let this fail silently and bubble up (currently)?
# XXX: local consumer has closed their side
# so cancel the far end streaming task
log.warning(f"{send_chan} consumer is already closed")
return False
# NOTE XXX: by default we do **not** maintain context-stream
# backpressure and instead opt to relay stream overrun errors to
# the sender; the main motivation is that using bp can block the
# msg handling loop which calls into this method!
except trio.WouldBlock:
# XXX: always push an error even if the local
# receiver is in overrun state.
# await self._maybe_cancel_and_set_remote_error(msg)
local_uid = current_actor().uid
lines = [
f'Actor-task context {cid}@{local_uid} was overrun by remote!',
f'sender actor: {uid}',
]
if not self._stream_opened:
lines.insert(
1,
f'\n*** No stream open on `{local_uid[0]}` side! ***\n'
)
text = '\n'.join(lines)
# XXX: lul, this really can't be backpressure since any
# blocking here will block the entire msg loop rpc sched for
# a whole channel.. maybe we should rename it?
if self._allow_overruns:
text += f'\nStarting overflow queuing task on msg: {msg}'
log.warning(text)
if (
not self._in_overrun
):
self._overflow_q.append(msg)
n = self._scope_nursery
if n.child_tasks:
from . import _debug
await _debug.breakpoint()
assert not n.child_tasks
n.start_soon(
self._drain_overflows,
)
else:
try:
raise StreamOverrun(text)
except StreamOverrun as err:
err_msg = pack_error(err)
err_msg['cid'] = cid
try:
await chan.send(err_msg)
except trio.BrokenResourceError:
# XXX: local consumer has closed their side
# so cancel the far end streaming task
log.warning(f"{chan} is already closed")
return False
def mk_context(
chan: Channel,
cid: str,
msg_buffer_size: int = 2**6,
**kwargs,
) -> Context:
'''
Internal factory to create an inter-actor task ``Context``.
This is called by internals and should generally never be called
by user code.
'''
send_chan: trio.MemorySendChannel
recv_chan: trio.MemoryReceiveChannel
send_chan, recv_chan = trio.open_memory_channel(msg_buffer_size)
ctx = Context(
chan,
cid,
_send_chan=send_chan,
_recv_chan=recv_chan,
**kwargs,
)
ctx._result = id(ctx)
return ctx
def stream(func: Callable) -> Callable:
'''