goodboy
/
tractor
1
0
Fork 2
Code Issues 5 Pull Requests 6 Packages Projects Releases Wiki Activity

Prefer fresh conn for unreg, fallback to `_parent_chan` 

The prior approach eagerly reused `_parent_chan` when
parent IS the registrar, but that channel may still
carry ctx/stream teardown protocol traffic —
concurrent `unregister_actor` RPC causes protocol
conflicts. Now try a fresh `get_registry()` conn
first; only fall back to the parent channel on
`OSError` (listener already closed/unlinked).

Deats,
- fresh `get_registry()` is the primary path for
  all addrs regardless of `parent_is_reg`
- `OSError` handler checks `parent_is_reg` +
  `rent_chan.connected()` before fallback
- fallback catches `OSError` and
  `trio.ClosedResourceError` separately
- drop unused `reg_addr: Address` annotation

(this commit msg was generated in some part by [`claude-code`][claude-code-gh])
[claude-code-gh]: https://github.com/anthropics/claude-code
subint_spawner_backend
Gud Boi 2026-04-14 16:51:48 -04:00
parent 70dc60a199
commit 8817032c90
1 changed files with 36 additions and 18 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

54
tractor/runtime/_runtime.py
View File

@ -1848,17 +1848,21 @@ async def async_main(
failed_unreg: bool = False failed_unreg: bool = False
rent_chan: Channel|None = actor._parent_chan rent_chan: Channel|None = actor._parent_chan
# XXX check if the parent IS the registrar so we # XXX, detect whether the parent IS the registrar
# can reuse the existing `_parent_chan` (avoids # so we can FALL BACK to `_parent_chan` when a new
# opening a new connection which fails when the # connection attempt fails (e.g. UDS transport
# listener socket is already closed, e.g. UDS # `os.unlink()`s the socket file during teardown).
# transport `os.unlink()`s the socket file during #
# teardown). # IMPORTANT: we do NOT eagerly reuse `_parent_chan`
# because it may still be carrying context/stream
# teardown protocol traffic — sending an
# `unregister_actor` RPC over it concurrently
# causes protocol-level conflicts. Instead we try
# a fresh `get_registry()` connection first and
# only fall back to the parent channel on failure.
# #
# See `ipc._uds.close_listener()` for details on # See `ipc._uds.close_listener()` for details on
# the UDS socket-file lifecycle and why this # the UDS socket-file lifecycle.
# optimization is necessary for the local-registrar
# case.
parent_is_reg: bool = False parent_is_reg: bool = False
if ( if (
rent_chan is not None rent_chan is not None
@ -1867,7 +1871,6 @@ async def async_main(
): ):
pchan_raddr: Address|None = rent_chan.raddr pchan_raddr: Address|None = rent_chan.raddr
if pchan_raddr is not None: if pchan_raddr is not None:
reg_addr: Address
for reg_addr in actor.reg_addrs: for reg_addr in actor.reg_addrs:
if ( if (
pchan_raddr.unwrap() pchan_raddr.unwrap()
@ -1883,24 +1886,39 @@ async def async_main(
with trio.move_on_after(0.5) as cs: with trio.move_on_after(0.5) as cs:
cs.shield = True cs.shield = True
try: try:
if parent_is_reg: async with get_registry(
reg_portal = Portal(rent_chan) addr,
) as reg_portal:
await reg_portal.run_from_ns( await reg_portal.run_from_ns(
'self', 'self',
'unregister_actor', 'unregister_actor',
uid=actor.aid.uid, uid=actor.aid.uid,
) )
else: except OSError:
async with get_registry( # Connection to registrar failed
addr, # (listener socket likely already
) as reg_portal: # closed/unlinked). Fall back to
# parent channel if parent IS the
# registrar.
if (
parent_is_reg
and
rent_chan.connected()
):
try:
reg_portal = Portal(rent_chan)
await reg_portal.run_from_ns( await reg_portal.run_from_ns(
'self', 'self',
'unregister_actor', 'unregister_actor',
uid=actor.aid.uid, uid=actor.aid.uid,
) )
except OSError: except (
failed_unreg = True OSError,
trio.ClosedResourceError,
):
failed_unreg = True
else:
failed_unreg = True
if cs.cancelled_caught: if cs.cancelled_caught:
failed_unreg = True failed_unreg = True