jprdonnelly/impala
1
0
Fork 0
mirror of https://github.com/apache/impala.git synced 2025-12-19 18:12:08 -05:00
Code Issues Projects Releases Wiki Activity

IMPALA-14271: Reapply the core piece of IMPALA-6984

Browse Source 
IMPALA-6984 changed the behavior to cancel backends when the
query reaches the RETURNED_RESULTS state. This ran into a regression
on large clusters where a query would end up waiting 10 seconds.
IMPALA-10047 reverted the core piece of the change.

For tuple caching, we found that a scan node can get stuck waiting
for a global runtime filter. It turns out that the coordinator will
not send out global runtime filters if the query is in a terminal
state. Tuple caching was causing queries to reach the RETURNED_RESULTS
phase before the runtime filter could be sent out. Reenabling the core
part of IMPALA-6984 sends out a cancel as soon as the query transitions
to RETURNED_RESULTS and wakes up any fragment instances waiting on
runtime filters.

The underlying cause of IMPALA-10047 is a tangle of locks that causes
us to exhaust the RPC threads. The coordinator is holding a lock on the
backend state while it sends the cancel synchronously. Other backends
that complete during that time run Coordinator::BackendState::LogFirstInProgress(),
which iterates through backend states to find the first that is not done.
The check to see if a backend state is done takes a lock on the backend
state. The problem case is that the coordinator may be sending a cancel
to a backend on itself. In that case, it needs an RPC thread on the coordinator
to be available to process the cancel. If all of the RPC threads are
processing updates, they can all call LogFirstInProgress() and get stuck
on the backend state lock for the coordinator's fragment. In that case,
it becomes a temporary deadlock as the cancel can't be processed and the
coordinator won't release the lock. It only gets resolved by the RPC timing
out.

To resolve this, this changes the Cancel() method to drop the lock while
doing the CancelQueryFInstances RPC. It reacquires the lock when it finishes
the RPC.

Testing:
 - Hand tested with 10 impalads and control_service_num_svc_threads=1
   Without the fix, it reproduces easily after reverting IMPALA-10047.
   With the fix, it doesn't reproduce.

Change-Id: Ia058b03c72cc4bb83b0bd0a19ff6c8c43a647974
Reviewed-on: http://gerrit.cloudera.org:8080/23264
Reviewed-by: Yida Wu <wydbaggio000@gmail.com>
Reviewed-by: Michael Smith <michael.smith@cloudera.com>
Tested-by: Impala Public Jenkins <impala-public-jenkins@cloudera.com>
This commit is contained in:
Joe McDonnell
2025-08-06 23:27:42 -07:00
committed by Impala Public Jenkins
parent e5afebc0c1
commit 16c350ce5a
2 changed files with 22 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

19
be/src/runtime/coordinator-backend-state.cc
View File

@@ -697,6 +697,13 @@ Coordinator::BackendState::CancelResult Coordinator::BackendState::Cancel(
goto done;
}
// Drop the lock while we send the cancel RPC. RPC threads from ReportExecStatus
// acquire the lock when calling IsDone() via LogFirstInProgress(). If this holds the
// lock for an extended period of time, it can block all of the RPC threads, leaving
// no thread to process this CancelQueryFInstances RPC locally. This causes a
// temporary deadlock until this RPC times out (10 seconds).
l.unlock();
CancelQueryFInstancesRequestPB request;
*request.mutable_query_id() = query_id_;
CancelQueryFInstancesResponsePB response;
@@ -709,6 +716,18 @@ Coordinator::BackendState::CancelResult Coordinator::BackendState::Cancel(
request, &response, query_ctx_, "Cancel() RPC failed", num_retries,
timeout_ms, backoff_time_ms, "COORD_CANCEL_QUERY_FINSTANCES_RPC");
// Reacquire the lock after the RPC completes
l.lock();
// If this became done while we dropped the lock for the CancelQueryFInstances RPC,
// the results of the RPC should be ignored.
if (IsDoneLocked(l)) {
VLogForBackend(Substitute(
"Ignoring result of cancel RPC because the backend is already done: $0",
status_.GetDetail()));
goto done;
}
if (!rpc_status.ok()) {
status_.MergeStatus(rpc_status);
result.became_done = true;

4
be/src/runtime/coordinator.cc
View File

@@ -872,7 +872,9 @@ void Coordinator::HandleExecStateTransition(
// execution and release resources.
ReleaseExecResources();
if (new_state == ExecState::RETURNED_RESULTS) {
// TODO: IMPALA-6984: cancel all backends in this case too.
// Cancel all backends, but wait for the final status reports to be received so that
// we have a complete profile for this successful query.
CancelBackends(/*fire_and_forget=*/ false);
WaitForBackends();
} else {
CancelBackends(/*fire_and_forget=*/ true);