Commit e070145d authored by jan.koester's avatar jan.koester
Browse files

cluster: add temporary [SESSIONS-DIAG] logging to listAllSessions() and push_worker_loop() 

Both paths currently swallow every failure silently (catch(...) {}, and
push_worker_loop only logs a drop after 5 retries, never a success) so
there's no way to tell from logs whether session groups are missing,
retrieve()/deserialize is failing, or the write side never lands.
Remove once the empty-admin-session-list issue is root-caused.
parent 8f58479c

src/cluster.cpp

+36 −6
Changes for src/cluster.cpp: 36 added lines, 6 removed lines.
Original line number Diff line number Diff line
@@ -775,7 +775,14 @@ namespace authdb { 
            if (!running_) break;



            // Don't try to push while critical

            if (critical_) continue;

            if (critical_) {

                if (!push_queue_.empty() || !data_push_queue_.empty()) {

                    std::cerr << "[SESSIONS-DIAG] push_worker_loop: skipped, cluster critical, "

                              << "pending_sessions=" << push_queue_.size()

                              << " pending_data=" << data_push_queue_.size() << std::endl;

                }

                continue;

            }



            // Drain ALL pending session pushes FIRST (more time-critical)

            std::vector<PendingPush> sess_batch;
@@ -797,6 +804,8 @@ namespace authdb { 
            for (auto &item : sess_batch) {

                try {

                    session_client_->store(item.sid_gid, item.data.data(), item.data.size());

                    std::cerr << "[SESSIONS-DIAG] push_worker: stored session gid=" << item.sid_gid

                              << " data_size=" << item.data.size() << std::endl;

                } catch (const std::exception &e) {

                    item.retries++;

                    if (item.retries < 5) {
@@ -972,12 +981,17 @@ namespace authdb { 
        std::set<uint64_t> all_groups;

        try {

            size_t n = session_read_client_->get_cluster_status().nodes_total;

            std::cerr << "[SESSIONS-DIAG] listAllSessions: nodes_total=" << n << std::endl;

            for (size_t i = 0; i < n; i++) {

                std::vector<uint64_t> node_groups;

                if (session_read_client_->list_groups_on_node(i, node_groups)) {

                bool ok = session_read_client_->list_groups_on_node(i, node_groups);

                std::cerr << "[SESSIONS-DIAG] list_groups_on_node(" << i << ") ok=" << ok

                          << " groups=" << node_groups.size() << std::endl;

                if (ok) {

                    all_groups.insert(node_groups.begin(), node_groups.end());

                }

            }

            std::cerr << "[SESSIONS-DIAG] all_groups (union) size=" << all_groups.size() << std::endl;



            // Try to retrieve and deserialize each group as a session

            for (uint64_t gid : all_groups) {
@@ -990,20 +1004,36 @@ namespace authdb { 
                    // session in the cluster instead of failing that group closed.

                    auto deadline = std::chrono::steady_clock::now() + std::chrono::seconds(5);

                    auto data = session_read_client_->retrieve(gid, deadline);

                    std::cerr << "[SESSIONS-DIAG] retrieve gid=" << gid

                              << " data_size=" << data.size() << std::endl;

                    if (data.empty()) continue;

                    SessionInfo info;

                    if (SessionBlock::deserialize(data, info.sid, info.uid, info.did,

                    bool deser = SessionBlock::deserialize(data, info.sid, info.uid, info.did,

                                                  info.members, info.username, info.gpo_results,

                                                  info.created_at)) {

                                                  info.created_at);

                    std::cerr << "[SESSIONS-DIAG] deserialize gid=" << gid

                              << " ok=" << deser

                              << " sid=" << (deser ? info.sid.c_str() : "?") << std::endl;

                    if (deser) {

                        std::string key = info.sid.c_str();

                        if (!info.sid.empty() && seen_sids.find(key) == seen_sids.end()) {

                            seen_sids.insert(key);

                            out.push_back(std::move(info));

                        }

                    }

                } catch (...) {}

                } catch (const std::exception &e) {

                    std::cerr << "[SESSIONS-DIAG] retrieve/deserialize gid=" << gid

                              << " EXCEPTION: " << e.what() << std::endl;

                } catch (...) {

                    std::cerr << "[SESSIONS-DIAG] retrieve/deserialize gid=" << gid

                              << " unknown exception" << std::endl;

                }

            }

        } catch (const std::exception &e) {

            std::cerr << "[SESSIONS-DIAG] listAllSessions outer EXCEPTION: " << e.what() << std::endl;

        } catch (...) {

            std::cerr << "[SESSIONS-DIAG] listAllSessions outer unknown exception" << std::endl;

        }

        } catch (...) {}

    }



    // --- DataOpBlock serialization ---