test (f1f2554f) · Commits · tuxist / authdb

src/cluster.cpp

+21 −0

Original line number	Diff line number	Diff line
		@@ -582,6 +582,8 @@ namespace authdb {
		size_t k = cfg_.data_blocks;
		size_t n = cfg_.data_blocks + cfg_.parity_blocks;
		bool first_run = true;
		int healthy_cycles = 0; // counts consecutive healthy probes
		static constexpr int REBALANCE_INTERVAL = 30; // rebalance every ~5min (30 × 10s)
		while (running_) {
		if (!first_run) {
		// Probe faster when degraded/critical (3s) vs normal (10s)
		@@ -643,9 +645,27 @@ namespace authdb {
		std::thread([this](){
		try { scrub(); } catch (...) {}
		}).detach();
		healthy_cycles = 0;
		}
		degraded_ = false;
		critical_ = false;

		// Periodic rebalance while healthy: fixes misplaced blocks
		// from transient store_stripe timeouts that didn't trigger
		// a full degraded→recovered transition.
		if (++healthy_cycles >= REBALANCE_INTERVAL) {
		healthy_cycles = 0;
		std::thread([this](){
		try {
		auto rb = pclient_->rebalance();
		if (rb.rebalanced > 0) {
		std::cerr << "Cluster: periodic rebalance moved "
		<< rb.rebalanced << " group(s)" << std::endl;
		pclient_->vacuum_all_nodes();
		}
		} catch (...) {}
		}).detach();
		}
		} else if (health.nodes_online >= k) {
		if (was_critical) {
		std::cerr << "Cluster: recovering — " << health.nodes_online
		@@ -656,6 +676,7 @@ namespace authdb {
		}
		degraded_ = true;
		critical_ = false;
		healthy_cycles = 0;
		} else {
		if (!critical_) {
		std::cerr << "Cluster: CRITICAL — only " << health.nodes_online