MB-47462: Expose CheckpointDestroyerTask mem usage metric

This metric allows monitoring of the memory used by checkpoints which
have been detached from their manager.

These checkpoints are still accounted for by EPStats
`estimatedCheckpointMemUsage`, but it is useful to directly identify
this memory usage.

Change-Id: I43011fe6ea2f7e8d66f04fc56ff372f5d1f6d139
Reviewed-on: http://review.couchbase.org/c/kv_engine/+/163354
Tested-by: Build Bot <[email protected]>
Reviewed-by: Dave Rigby <[email protected]>

Author: jameseh96

engines/ep/docs/stats.org

@@ -403,24 +403,26 @@ The following ratios are encoded as 4-digit integers, e.g.:
 
 ** vBucket total stats
 
-| Stat                              | Description                                    |
-|-----------------------------------+------------------------------------------------|
-| ep_vb_total                       | Total vBuckets (count)                         |
-| curr_items_tot                    | Total number of items                          |
-| curr_items                        | Number of active items in memory               |
-| curr_temp_items                   | Number of temporary items in memory            |
-| vb_dead_num                       | Number of dead vBuckets                        |
-| ep_diskqueue_items                | Total items in disk queue                      |
-| ep_diskqueue_memory               | Total memory used in disk queue                |
-| ep_diskqueue_fill                 | Total enqueued items on disk queue             |
-| ep_diskqueue_drain                | Total drained items on disk queue              |
-| ep_diskqueue_pending              | Total bytes of pending writes                  |
-| ep_persist_vbstate_total          | Total VB persist state to disk                 |
-| ep_meta_data_memory               | Total memory used by meta data                 |
-| ep_meta_data_disk                 | Total disk used by meta data                   |
-| ep_checkpoint_memory              | Memory of items in all checkpoints             |
-| ep_checkpoint_memory_unreferenced | Memory of items in unref checkpoints           |
-| ep_checkpoint_memory_overhead     | Memory of all checkpoints structures           |
+| Stat                                     | Description                                    |
+|------------------------------------------+------------------------------------------------|
+| ep_vb_total                              | Total vBuckets (count)                         |
+| curr_items_tot                           | Total number of items                          |
+| curr_items                               | Number of active items in memory               |
+| curr_temp_items                          | Number of temporary items in memory            |
+| vb_dead_num                              | Number of dead vBuckets                        |
+| ep_diskqueue_items                       | Total items in disk queue                      |
+| ep_diskqueue_memory                      | Total memory used in disk queue                |
+| ep_diskqueue_fill                        | Total enqueued items on disk queue             |
+| ep_diskqueue_drain                       | Total drained items on disk queue              |
+| ep_diskqueue_pending                     | Total bytes of pending writes                  |
+| ep_persist_vbstate_total                 | Total VB persist state to disk                 |
+| ep_meta_data_memory                      | Total memory used by meta data                 |
+| ep_meta_data_disk                        | Total disk used by meta data                   |
+| ep_checkpoint_memory                     | Memory of items in all checkpoints             |
+| ep_checkpoint_memory_unreferenced        | Memory of items in unref checkpoints           |
+| ep_checkpoint_memory_overhead            | Memory of all checkpoints structures           |
+| ep_checkpoint_memory_pending_destruction | Memory of checkpoint structures awaiting       |
+|                                          | destruction by a background task               |
 
 *** Active vBucket class stats
 

engines/ep/src/ep_engine.cc

@@ -3107,6 +3107,9 @@ cb::engine_errc EventuallyPersistentEngine::doEngineStatsLowCardinality(
     kvBucket->getAggregatedVBucketStats(collector,
                                         cb::prometheus::Cardinality::Low);
 
+    collector.addStat(Key::ep_checkpoint_memory_pending_destruction,
+                      kvBucket->getCheckpointPendingDestructionMemoryUsage());
+
     kvBucket->getFileStats(collector);
 
     collector.addStat(Key::ep_persist_vbstate_total,

engines/ep/src/kv_bucket.cc

@@ -2958,3 +2958,11 @@ CheckpointDestroyerTask& KVBucket::getCkptDestroyerTask(Vbid vbid) const {
     Expects(!ckptDestroyerTasks.empty());
     return *ckptDestroyerTasks[vbid.get() % ckptDestroyerTasks.size()];
 }
+
+size_t KVBucket::getCheckpointPendingDestructionMemoryUsage() const {
+    size_t memoryUsage = 0;
+    for (const auto& task : ckptDestroyerTasks) {
+        memoryUsage += task->getMemoryUsage();
+    }
+    return memoryUsage;
+}

engines/ep/src/kv_bucket.h

@@ -878,6 +878,8 @@ class KVBucket : public KVBucketIface {
         return chkRemovers.size();
     }
 
+    size_t getCheckpointPendingDestructionMemoryUsage() const;
+
 protected:
     /**
      * Get the checkpoint destroyer task responsible for checkpoints from the

engines/ep/tests/ep_testsuite.cc

@@ -7164,6 +7164,7 @@ static enum test_result test_mb19687_fixed(EngineIface* h) {
               "ep_chk_persistence_remains",
               "ep_chk_remover_stime",
               "ep_checkpoint_destruction_tasks",
+              "ep_checkpoint_memory_pending_destruction",
               "ep_checkpoint_memory_ratio",
               "ep_checkpoint_memory_recovery_upper_mark",
               "ep_checkpoint_memory_recovery_lower_mark",

include/statistics/stats.def.h

@@ -584,6 +584,7 @@ STAT(ep_meta_data_disk, , bytes, , )
 STAT(ep_checkpoint_memory, , bytes, , )
 STAT(ep_checkpoint_memory_unreferenced, , bytes, , )
 STAT(ep_checkpoint_memory_overhead, , bytes, , )
+STAT(ep_checkpoint_memory_pending_destruction, , bytes, , )
 STAT(ep_total_cache_size, , bytes, , )
 STAT(rollback_item_count, , count, , )
 STAT(ep_num_non_resident, , count, , )