zbtrs2
/
Mooncake
forked from mooncake-track/Mooncake
0
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

change adxl log (#1039) 

Co-authored-by: youxiao <youxiao@huawei.com>
This commit is contained in:
ascend-direct-dev 2025-11-11 09:44:03 +08:00 committed by GitHub
parent 09503750d1
commit b336a68632
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
3 changed files with 25 additions and 6 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
mooncake-store/src/transfer_task.cpp
View File

@ -244,9 +244,11 @@ void TransferEngineOperationState::check_task_status() {
case TransferStatusEnum::FAILED:
case TransferStatusEnum::CANCELED:
case TransferStatusEnum::INVALID:
#ifndef USE_ASCEND_DIRECT
LOG(ERROR) << "Transfer failed for batch " << batch_id_
<< " task " << i << " with status "
<< static_cast<int>(status.s);
#endif
has_failure = true;
break;
default:

4
mooncake-transfer-engine/include/transport/ascend_transport/ascend_direct_transport/ascend_direct_transport.h
View File

@ -124,8 +124,8 @@ class AscendDirectTransport : public Transport {
int32_t device_logic_id_{};
aclrtContext rt_context_{nullptr};
int32_t connect_timeout_ = 3000;
int32_t transfer_timeout_ = 3000;
int32_t connect_timeout_ = 10000;
int32_t transfer_timeout_ = 10000;
std::string local_adxl_engine_name_{};
aclrtStream stream_{};
bool use_buffer_pool_{false};

25
mooncake-transfer-engine/src/transport/ascend_transport/ascend_direct_transport/ascend_direct_transport.cpp
View File

@ -573,8 +573,14 @@ void AscendDirectTransport::processSliceList(
return;
}
if (target_adxl_engine_name == local_adxl_engine_name_) {
VLOG(1) << "Target is local, use memory copy.";
return localCopy(slice_list[0]->opcode, slice_list);
auto start = std::chrono::steady_clock::now();
localCopy(slice_list[0]->opcode, slice_list);
LOG(INFO) << "Local copy time: "
<< std::chrono::duration_cast<std::chrono::microseconds>(
std::chrono::steady_clock::now() - start)
.count()
<< "us";
return;
}
int ret = checkAndConnect(target_adxl_engine_name);
if (ret != 0) {
@ -608,7 +614,14 @@ void AscendDirectTransport::processSliceList(
.count()
<< " us";
} else {
LOG(ERROR) << "Transfer slice failed with status: " << status;
if (status == adxl::TIMEOUT) {
LOG(ERROR) << "Transfer timeout to: " << target_adxl_engine_name
<< ", you can increase the timeout duration to reduce "
"the failure rate by configuring "
"the ASCEND_TRANSFER_TIMEOUT environment variable.";
} else {
LOG(ERROR) << "Transfer slice failed with status: " << status;
}
for (auto &slice : slice_list) {
slice->markFailed();
}
@ -816,7 +829,11 @@ int AscendDirectTransport::checkAndConnect(
}
auto status =
adxl_->Connect(target_adxl_engine_name.c_str(), connect_timeout_);
if (status != adxl::SUCCESS) {
if (status == adxl::TIMEOUT) {
LOG(ERROR) << "Connect timeout to: " << target_adxl_engine_name
<< ", you can increase the timeout duration to reduce "
"the ASCEND_CONNECT_TIMEOUT environment variable.";
} else if (status != adxl::SUCCESS) {
LOG(ERROR) << "Failed to connect to target: " << target_adxl_engine_name
<< ", status: " << status;
return -1;