Skip to content

bugfix: modify the link/unlink workflow for MIX instances. #47

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
JimHsiung merged 1 commit into from
Mar 18, 2026
Merged

bugfix: modify the link/unlink workflow for MIX instances. #47

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion xllm_service/common/time_predictor.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -70,7 +70,7 @@ TimePredictor::TimePredictor(
// get coefficients
tpot_coefficients_ = matrix.colPivHouseholderQr().solve(target);
} else {
ttft_coefficients_ = Eigen::VectorXd::Zero(3);
tpot_coefficients_ = Eigen::VectorXd::Zero(3);
}
}

Expand Down
46 changes: 18 additions & 28 deletions xllm_service/scheduler/managers/instance_mgr.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -795,6 +795,7 @@ bool InstanceMgr::link_instance_internal(const std::string& name,
bool link_ok = true;
int32_t linked_p_count = 0;
int32_t linked_d_count = 0;
std::vector<std::string> linked_mix_names;

switch (info.type) {
case InstanceType::DEFAULT:
Expand All @@ -820,23 +821,15 @@ bool InstanceMgr::link_instance_internal(const std::string& name,
break;
}
case InstanceType::MIX: {
for (auto& p_name : prefill_index_) {
if (!call_link_instance(info.rpc_address, instances_[p_name]) ||
!call_link_instance(instances_[p_name].rpc_address, info)) {
for (const auto& [peer_name, peer_info] : instances_) {
if (peer_name == name) {
continue;
}
if (!call_link_instance(info.rpc_address, peer_info)) {
Comment thread
JimHsiung marked this conversation as resolved.
link_ok = false;
break;
}
linked_p_count++;
}
if (link_ok) {
for (auto& d_name : decode_index_) {
if (!call_link_instance(info.rpc_address, instances_[d_name]) ||
!call_link_instance(instances_[d_name].rpc_address, info)) {
link_ok = false;
break;
}
linked_d_count++;
}
linked_mix_names.emplace_back(peer_name);
}
break;
}
Expand All @@ -858,13 +851,12 @@ bool InstanceMgr::link_instance_internal(const std::string& name,
call_unlink_instance(info.rpc_address, instances_[prefill_index_[i]]);
}
} else if (info.type == InstanceType::MIX) {
for (int i = 0; i < linked_p_count; i++) {
call_unlink_instance(info.rpc_address, instances_[prefill_index_[i]]);
call_unlink_instance(instances_[prefill_index_[i]].rpc_address, info);
}
for (int i = 0; i < linked_d_count; i++) {
call_unlink_instance(info.rpc_address, instances_[decode_index_[i]]);
call_unlink_instance(instances_[decode_index_[i]].rpc_address, info);
for (const auto& linked_name : linked_mix_names) {
auto it = instances_.find(linked_name);
if (it == instances_.end()) {
continue;
}
call_unlink_instance(info.rpc_address, it->second);
Comment thread
JimHsiung marked this conversation as resolved.
}
}
return false;
Expand All @@ -884,13 +876,11 @@ void InstanceMgr::unlink_instance_internal(const std::string& name,
call_unlink_instance(instances_[p_name].rpc_address, info);
}
} else if (info.type == InstanceType::MIX) {
for (auto& p_name : prefill_index_) {
if (p_name == name) continue;
call_unlink_instance(instances_[p_name].rpc_address, info);
}
for (auto& d_name : decode_index_) {
if (d_name == name) continue;
call_unlink_instance(instances_[d_name].rpc_address, info);
for (const auto& [peer_name, peer_info] : instances_) {
if (peer_name == name) {
continue;
}
call_unlink_instance(peer_info.rpc_address, info);
}
}
}
Expand Down
Loading