fix: Flush pool as late as possible during schema cache reloading

retryingSchemaCacheLoad flushes the pool upon every retry before it starts reloading the schema. This is too early as schema reloading might take some time during which new connections might be acquired. The consequence is that:
* upon successful schema cache reload we might have some connections created with the old schema cache
* we close connections upon each retry and under load we will keep closing and re-opening connections until schema cache load succeeds

This change is to make sure we flush the pool only after successful schema cache querying but before loading (so that connections acquired during loading wait for it and do not interfere with timing the loading process).

Author: mkleczek

src/PostgREST/AppState.hs

@@ -371,8 +371,6 @@ retryingSchemaCacheLoad appState@AppState{stateObserver=observer, stateMainThrea
       observer $ ConnectionRetryObs delay
       putNextListenerDelay appState delay
 
-    flushPool appState
-
     (,) <$> qPgVersion <*> (qInDbConfig *> qSchemaCache)
   )
   where
@@ -421,6 +419,10 @@ retryingSchemaCacheLoad appState@AppState{stateObserver=observer, stateMainThrea
           -- IORef on putSchemaCache. This is why SCacheStatus is put at SCPending here to signal the Admin server (using isPending) that we're on a recovery state.
           putSCacheStatus appState SCPending
           putSchemaCache appState $ Just sCache
+          -- Flush the pool after loading the schema cache to reset any stale session cache entries
+          -- We do it after successfully querying the schema cache
+          -- and after marking sCacheStatus as pending,
+          flushPool appState
           observer $ SchemaCacheQueriedObs resultTime
           (t, _) <- timeItT $ observer $ SchemaCacheSummaryObs $ showSummary sCache
           observer $ SchemaCacheLoadedObs t

test/io/test_io.py

@@ -674,7 +674,7 @@ def test_log_level(level, defaultenv):
         response = postgrest.session.get("/")
         assert response.status_code == 200
 
-        output = postgrest.read_stdout(nlines=7)
+        output = postgrest.read_stdout(nlines=9)
 
         if level == "crit":
             assert len(output) == 0
@@ -712,7 +712,7 @@ def test_log_level(level, defaultenv):
                     r'- - postgrest_test_anonymous \[.+\] "GET / HTTP/1.1" 200 \d+ "" "python-requests/.+"',
                 ],
             )
-            assert len(output) == 7
+            assert len(output) == 9
             assert any("Connection" and "is available" in line for line in output)
             assert any("Connection" and "is used" in line for line in output)
 
@@ -1349,7 +1349,7 @@ def test_db_error_logging_to_stderr(level, defaultenv, metapostgrest):
         assert response.status_code == 500
 
         # ensure the message appears on the logs
-        output = postgrest.read_stdout(nlines=6)
+        output = postgrest.read_stdout(nlines=8)
 
         if level == "crit":
             assert len(output) == 0
@@ -1566,7 +1566,7 @@ def test_log_pool_req_observation(level, defaultenv):
 
         if level == "debug":
             output = postgrest.read_stdout(nlines=7)
-            assert len(output) == 6
+            assert len(output) == 7
             match_log(output, [pool_req, pool_req_fullfill])
         elif level == "info":
             output = postgrest.read_stdout(nlines=4)

test/spec/Feature/SchemaCacheSpec.hs

@@ -49,22 +49,21 @@ spec = describe "Server started with metrics enabled" $ do
       waitFor (1 * sec) "SchemaCacheLoadedObs" $ \x -> [ o | o@SchemaCacheLoadedObs{} <- pure x ]
 
 
-  it "Should flush pool multiple times when schema reloading retries" $ do
+  it "Should flush pool once when schema reloading retries" $ do
     (appState, waitFor) <- prepareState
 
     liftIO $ do
         AppState.getConfig appState >>= \cfg -> do
           AppState.putConfig appState $ cfg { configDbSchemas = pure "bad_schema" }
           AppState.schemaCacheLoader appState
 
-          waitFor (1 * sec) "PoolFlushed 1" $ \x -> [ o | o@PoolFlushed <- pure x ]
           waitFor (1 * sec) "SchemaCacheErrorObs" $ \x -> [ o | o@SchemaCacheErrorObs{} <- pure x ]
 
           -- Restore configuration
           AppState.putConfig appState cfg
 
         -- Wait for 2 seconds so that retry can happen
-        waitFor (2 * sec) "PoolFlushed 2" $ \x -> [ o | o@PoolFlushed <- pure x ]
+        waitFor (2 * sec) "PoolFlushed" $ \x -> [ o | o@PoolFlushed <- pure x ]
         waitFor (1 * sec) "SchemaCacheQueriedObs" $ \x -> [ o | o@SchemaCacheQueriedObs{} <- pure x ]
         waitFor (1 * sec) "SchemaCacheLoadedObs" $ \x -> [ o | o@SchemaCacheLoadedObs{} <- pure x ]