Commit a310c53e authored by jan.koester's avatar jan.koester
Browse files

test

parent 5aae08e0

src/backend.cpp

+9 −25
Original line number Diff line number Diff line
@@ -18,10 +18,10 @@ namespace mediadb { 


static int sync_interval_seconds() {

    const char* v = std::getenv("MEDIADB_SYNC_INTERVAL_SEC");

    if (!v || v[0] == '\0') return 5;

    if (!v || v[0] == '\0') return 60;

    int x = std::atoi(v);

    if (x < 1) x = 1;

    if (x > 300) x = 300;

    if (x > 3600) x = 3600;

    return x;

}
@@ -3352,19 +3352,6 @@ void ClusterMediaBackend::sync_from_cluster_now() { 
        return;

    }



    uint64_t index_gid = cluster_group_id("index");

    bool has_index = false;

    for (const auto& pg : cluster_.list_peer_groups()) {

        if (std::find(pg.groups.begin(), pg.groups.end(), index_gid) != pg.groups.end()) {

            has_index = true;

            break;

        }

    }



    // Warm up read connections before fetching (QUIC idle timeout may have

    // closed them since start() or last sync cycle)

    cluster_.warmup_read_clients();



    // Fetch index from cluster (contains store/album/ACL metadata, no media data)

    std::vector<uint8_t> index_data;

    bool fetch_ok = cluster_.fetch("index", index_data);
@@ -3377,8 +3364,7 @@ void ClusterMediaBackend::sync_from_cluster_now() { 
            std::cerr << "[CLUSTER-SYNC] import active, skipping index fetch retry\n";

            return;

        }

        // Retry once after warmup — connections may be stale after heavy

        // import replication traffic on peers.

        // Retry once after warmup — connections may be stale

        cluster_.warmup_read_clients();

        std::this_thread::sleep_for(std::chrono::milliseconds(500));

        fetch_ok = cluster_.fetch("index", index_data);
@@ -3392,13 +3378,12 @@ void ClusterMediaBackend::sync_from_cluster_now() { 
            std::cerr << "[CLUSTER-SYNC] import active after index fetch failure, suppressing sync error\n";

            return;

        }

        if (!has_index) {

            std::cerr << "[CLUSTER-SYNC] no index exists on any peer — fresh/empty cluster\n";

        if (fetch_ok) {

            // fetch succeeded but returned empty data — fresh/empty cluster

            std::cerr << "[CLUSTER-SYNC] index returned empty data — fresh/empty cluster\n";

            initial_sync_ok_.store(true);

        } else {

            std::cerr << "[CLUSTER-SYNC] index fetch "

                      << (fetch_ok ? "returned empty data" : "failed (exception in cluster layer)")

                      << " (has_index=true, peers report having group)\n";

            std::cerr << "[CLUSTER-SYNC] index fetch failed (exception in cluster layer)\n";

        }

        return;

    }
@@ -3450,8 +3435,7 @@ void ClusterMediaBackend::sync_from_cluster_now() { 
        std::vector<uint8_t> store_data;

        bool ok = cluster_.fetch("store:" + sid, store_data);

        if (!ok || store_data.empty()) {

            // Retry once after warmup — peer connections may be stale

            cluster_.warmup_read_clients();

            // Retry once — peer connections may be stale

            std::this_thread::sleep_for(std::chrono::milliseconds(300));

            ok = cluster_.fetch("store:" + sid, store_data);

        }
@@ -3835,7 +3819,7 @@ void ClusterMediaBackend::sync_loop() { 
            const auto now = std::chrono::steady_clock::now();

            if (now >= next_sync) {

                sync_requested_ = true;

                if (++cycle >= 6) {

                if (++cycle >= 10) {

                    cycle = 0;

                    repair_requested_ = true;

                }

src/cluster.cpp

+3 −10
Original line number Diff line number Diff line
@@ -549,7 +549,7 @@ void Cluster::health_loop() { 
    size_t k = cfg_.data_blocks;

    size_t n = cfg_.data_blocks + cfg_.parity_blocks;

    int healthy_cycles = 0;

    static constexpr int REBALANCE_INTERVAL = 10; // rebalance every ~5min (10 × 30s)

    static constexpr int REBALANCE_INTERVAL = 10; // rebalance every ~10min (10 × 60s)

    while (running_) {

        if (pclient_) {

            uint32_t online = 0;
@@ -564,13 +564,6 @@ void Cluster::health_loop() { 
                std::vector<uint64_t> dummy;

                if (pclient_->list_groups_on_node(it->second, dummy)) {

                    ++online;

                } else {

                    // Retry once after short delay — peers may be busy

                    // during import or scrub and miss the first probe

                    std::this_thread::sleep_for(std::chrono::milliseconds(500));

                    dummy.clear();

                    if (pclient_->list_groups_on_node(it->second, dummy))

                        ++online;

                }

            }

            server_->set_peer_status_callback([online, total]() -> std::pair<uint32_t, uint32_t> {
@@ -682,8 +675,8 @@ void Cluster::health_loop() { 
                critical_ = true;

            }

        }

        // Probe faster when degraded/critical (5s) vs normal (30s)

        int interval = (degraded_ || critical_) ? 5 : 30;

        // Probe faster when degraded/critical (10s) vs normal (60s)

        int interval = (degraded_ || critical_) ? 10 : 60;

        for (int i = 0; i < interval && running_; ++i)

            std::this_thread::sleep_for(std::chrono::seconds(1));

    }