add blk metrics for gc (#385)

Author: JacksonYao287

conanfile.py

@@ -10,7 +10,7 @@
 
 class HomeObjectConan(ConanFile):
     name = "homeobject"
-    version = "3.0.16"
+    version = "3.0.17"
 
     homepage = "https://github.com/eBay/HomeObject"
     description = "Blob Store built on HomeStore"

src/lib/homestore_backend/gc_manager.cpp

@@ -774,9 +774,6 @@ bool GCManager::pdev_gc_actor::copy_valid_data(
             */
         }
 #endif
-
-        // for emergent gc, we directly use the current shard header as the new header
-
         // TODO::involve ratelimiter in the following code, where read/write are scheduled. or do we need a central
         // ratelimter shared by all components except client io?
         auto succeed_copying_shard =
@@ -786,6 +783,8 @@ bool GCManager::pdev_gc_actor::copy_valid_data(
                             &valid_blob_indexes, &data_service, task_id, &last_shard_state, &copied_blobs, pg_id,
                             header_sgs = std::move(header_sgs)](auto&& err) {
                     RELEASE_ASSERT(header_sgs.iovs.size() == 1, "header_sgs.iovs.size() should be 1, but not!");
+                    // shard header occupies one blk
+                    COUNTER_INCREMENT(metrics_, gc_write_blk_count, 1);
                     iomanager.iobuf_free(reinterpret_cast< uint8_t* >(header_sgs.iovs[0].iov_base));
                     if (err) {
                         GCLOGE(task_id, pg_id, shard_id,
@@ -822,6 +821,7 @@ bool GCManager::pdev_gc_actor::copy_valid_data(
                             data_service.async_read(pba, data_sgs, total_size)
                                 .thenValue([this, k, &hints, &move_from_chunk, &move_to_chunk, &data_service, task_id,
                                             pg_id, data_sgs = std::move(data_sgs), pba, &copied_blobs](auto&& err) {
+                                    COUNTER_INCREMENT(metrics_, gc_read_blk_count, pba.blk_count());
                                     RELEASE_ASSERT(data_sgs.iovs.size() == 1,
                                                    "data_sgs.iovs.size() should be 1, but not!");
 
@@ -865,6 +865,7 @@ bool GCManager::pdev_gc_actor::copy_valid_data(
                                     return data_service.async_alloc_write(data_sgs, hints, new_pba)
                                         .thenValue([this, shard_id, blob_id, new_pba, &move_to_chunk, task_id, pg_id,
                                                     &copied_blobs, data_sgs = std::move(data_sgs)](auto&& err) {
+                                            COUNTER_INCREMENT(metrics_, gc_write_blk_count, new_pba.blk_count());
                                             RELEASE_ASSERT(data_sgs.iovs.size() == 1,
                                                            "data_sgs.iovs.size() should be 1, but not!");
                                             iomanager.iobuf_free(
@@ -937,8 +938,9 @@ bool GCManager::pdev_gc_actor::copy_valid_data(
                                 return folly::makeFuture< std::error_code >(std::error_code{});
                             }
 
-                            // write shard footer
+                            // write shard footer, which occupies one blk
                             homestore::MultiBlkId out_blkids;
+                            COUNTER_INCREMENT(metrics_, gc_write_blk_count, 1);
                             return data_service.async_alloc_write(footer_sgs, hints, out_blkids);
                         })
                         .thenValue([this, &move_to_chunk, &shard_id, footer_sgs, task_id, pg_id](auto&& err) {

src/lib/homestore_backend/gc_manager.hpp

@@ -132,6 +132,8 @@ class GCManager {
                 REGISTER_GAUGE(failed_egc_task_count, "Number of failed emergent gc tasks");
                 REGISTER_GAUGE(total_reclaimed_space_by_gc, "Total reclaimed space by gc task");
                 REGISTER_GAUGE(total_reclaimed_space_by_egc, "Total reclaimed space by emergent gc task");
+                REGISTER_COUNTER(gc_read_blk_count, "Total read blk count by gc in this pdev");
+                REGISTER_COUNTER(gc_write_blk_count, "Total written blk count by gc in this pdev");
 
                 // gc task level histogram metrics
                 REGISTER_HISTOGRAM(reclaim_ratio_gc, "the ratio of reclaimed blks to total blks in a gc task",
@@ -149,6 +151,7 @@ class GCManager {
                 register_me_to_farm();
                 attach_gather_cb(std::bind(&pdev_gc_metrics::on_gather, this));
             }
+
             ~pdev_gc_metrics() { deregister_me_from_farm(); }
             pdev_gc_metrics(const pdev_gc_metrics&) = delete;
             pdev_gc_metrics(pdev_gc_metrics&&) noexcept = delete;
@@ -313,7 +316,7 @@ class GCManager {
     void drain_pg_pending_gc_task(const pg_id_t pg_id);
     void decr_pg_pending_gc_task(const pg_id_t pg_id);
     void incr_pg_pending_gc_task(const pg_id_t pg_id);
-    auto& get_gc_actore_superblks() { return m_gc_actor_sbs; }
+    auto& get_gc_actor_superblks() { return m_gc_actor_sbs; }
     std::shared_ptr< pdev_gc_actor > get_pdev_gc_actor(uint32_t pdev_id);
 
 private:

src/lib/homestore_backend/hs_cp_callbacks.cpp

@@ -56,7 +56,7 @@ folly::Future< bool > HSHomeObject::MyCPCallbacks::cp_flush(CP* cp) {
 
     // flush gc durable_entities
     auto gc_manager = home_obj_.gc_manager();
-    auto& gc_actor_superblks = gc_manager->get_gc_actore_superblks();
+    auto& gc_actor_superblks = gc_manager->get_gc_actor_superblks();
     for (auto& gc_actor_sb : gc_actor_superblks) {
         const auto pdev_id = gc_actor_sb->pdev_id;
         const auto gc_actor = gc_manager->get_pdev_gc_actor(pdev_id);

src/lib/homestore_backend/hs_homeobject.hpp

@@ -479,10 +479,6 @@ class HSHomeObject : public HomeObjectImpl {
         homestore::MultiBlkId pbas;
     };
 
-    struct BlobInfoData : public BlobInfo {
-        Blob blob;
-    };
-
     enum class BlobState : uint8_t {
         ALIVE = 0,
         TOMBSTONE = 1,

src/lib/homestore_backend/replication_state_machine.cpp

@@ -724,14 +724,14 @@ folly::Future< std::error_code > ReplicationStateMachine::on_fetch_data(const in
                 auto rc = index_table->get(get_req);
                 if (sisl_unlikely(homestore::btree_status_t::success != rc)) {
                     // blob never exists or has been gc
-                    LOGD("on_fetch_data failed to get from index table, blob never exists or has been gc, blob_id={}, "
+                    LOGD("on_fetch_data: failed to get from index table, blob never exists or has been gc, blob_id={}, "
                          "shardID=0x{:x}, pg={}",
                          blob_id, shard_id, pg_id);
                     should_return_delete_marker = true;
                 } else {
                     pbas = index_value.pbas();
                     if (sisl_unlikely(pbas == HSHomeObject::tombstone_pbas)) {
-                        LOGD("on_fetch_data: blob has been deleted, blob_id={}, shardID=0x{:x}, pg={}", blob_id,
+                        LOGD("on_fetch_data: got tombstone pba for blob_id={}, shardID=0x{:x}, pg={}", blob_id,
                              shard_id, pg_id);
                         should_return_delete_marker = true;
                     }