scmstore: don't update ODS counters for every fetch

Summary:
Instead, update ODS counters during flush() (where the regular Sapling counters get updated). EdenFS calls flush() on its backing stores every 1 minute, so they should get updated regularly.

I'm doing this because the scmstore metrics weren't "designed" to be logged for every fetch. Getting the list of metrics allocates a ton of Strings, so calling for every fetch (which can often be a fetch of a single file or directory), is wasteful (didn't profile - could be neglible amount of waste).

Reviewed By: sggutier

Differential Revision: D66731805

fbshipit-source-id: 46578e2d18a6534c956ec71054b4bc2016e4c669

eden/scm/lib/revisionstore/src/scmstore/file.rs

@@ -339,9 +339,6 @@ impl FileStore {
             state.derive_computable(aux_cache.as_ref().map(|s| s.as_ref()));
 
             metrics.write().fetch += state.metrics().clone();
-            if let Err(err) = state.metrics().update_ods() {
-                tracing::error!("Error updating ods fetch metrics: {}", err);
-            }
             state.finish();
 
             if let Some(activity_logger) = activity_logger {
@@ -482,11 +479,13 @@ impl FileStore {
             aux_cache.flush().map_err(&mut handle_error);
         }
 
-        let mut metrics = self.metrics.write();
+        let metrics = std::mem::take(&mut *self.metrics.write());
         for (k, v) in metrics.metrics() {
             hg_metrics::increment_counter(k, v as u64);
         }
-        *metrics = Default::default();
+        if let Err(err) = metrics.fetch.update_ods() {
+            tracing::error!("Error updating ods fetch metrics: {}", err);
+        }
 
         result
     }

eden/scm/lib/revisionstore/src/scmstore/tree.rs

@@ -388,11 +388,6 @@ impl TreeStore {
 
             // TODO(meyer): Report incomplete / not found, handle errors better instead of just always failing the batch, etc
             state.common.results(state.errors);
-
-            if let Err(err) = state.metrics.update_ods() {
-                tracing::error!(?err, "error updating tree ods counters");
-            }
-
             store_metrics.write().fetch += state.metrics;
 
             Ok(())
@@ -477,11 +472,14 @@ impl TreeStore {
             historystore_cache.flush().map_err(&mut handle_error);
         }
 
-        let mut metrics = self.metrics.write();
+        let metrics = std::mem::take(&mut *self.metrics.write());
         for (k, v) in metrics.metrics() {
             hg_metrics::increment_counter(k, v as u64);
         }
-        *metrics = Default::default();
+
+        if let Err(err) = metrics.fetch.update_ods() {
+            tracing::error!(?err, "error updating tree ods counters");
+        }
 
         result
     }