Repository navigation
UCT/UD: Revert stale private connection cleanup in v1.23.x-fur - #12249
Conversation
|
🤖 Starting review — findings will be posted here when done. |
|
|
||
| if (ep->flags & UCT_UD_EP_FLAG_PRIVATE) { | ||
| uct_ud_ep_purge(ep, UCS_ERR_ENDPOINT_TIMEOUT); | ||
| ucs_assert(ucs_queue_is_empty(&ep->tx.window)); |
There was a problem hiding this comment.
uct_ud_ep_timer() reaches the peer-timeout flow only when tx.window is non-empty (e.g. a private ep still retransmitting an unacked CREP to a peer that is gone), so this assert can fail in debug builds. Can we keep the purge here?
| ucs_assert(ucs_queue_is_empty(&ep->tx.window)); | |
| uct_ud_ep_purge(ep, UCS_ERR_ENDPOINT_TIMEOUT); |
| void *ud_ep GTEST_ATTRIBUTE_UNUSED_; | ||
| return !ucs_ptr_array_lookup(&iface->eps, ep_id, ud_ep); | ||
| }, [this]() { short_progress_loop(); }, 60.0); | ||
| }, []() { usleep(1000); }, 60.0); |
There was a problem hiding this comment.
Why revert this hunk to a plain sleep? short_progress_loop() works for both remaining callers and additionally covers endpoints that are released from the worker progress callback.
| #include <ucs/vfs/base/vfs_obj.h> | ||
|
|
||
|
|
||
| /* Must be less then peer_timeout to avoid false positive errors taking into |
There was a problem hiding this comment.
Minor: then -> than
| /* Must be less then peer_timeout to avoid false positive errors taking into | |
| /* Must be less than peer_timeout to avoid false positive errors taking into |
|
Residual coverage note: the revert removes |
|
/azp run |
What?
Revert
9aa2c33e58e40306244461117a56af23458f7bac(#11899) and its dependent follow-up changes from027a92f7cd94d6ea35eb2214f4fb21b7a4674835, restoring the previous private UD endpoint lifecycle.The change is based directly on
v1.23.x-furat the HPC-X 2.52 UCX pin,8094f699daa0d53ea84d11b6cefd7dffa9c9eb46.Why?
Address MPI/UCC collective hangs caused by stale private-endpoint cleanup while retaining the HPC-X pinned baseline.