SERVER-34338 Find explain in legacy query path on mongos does not retry on StaleShardVersion

(cherry picked from commit 3e31679dfa6d85ebe48855582304c8cb7a635b0d)

1 files changed, 68 insertions, 23 deletions

diff --git a/src/mongo/s/commands/strategy.cpp b/src/mongo/s/commands/strategy.cpp
index 80873b48382..89d5affdb1d 100644
--- a/src/mongo/s/commands/strategy.cpp
+++ b/src/mongo/s/commands/strategy.cpp
@@ -362,7 +362,7 @@ void runCommand(OperationContext* opCtx,
             try {
                 execCommandClient(opCtx, invocation.get(), request, &crb);
                 return;
-            } catch (ExceptionForCat<ErrorCategory::NeedRetargettingError>& ex) {
+            } catch (const ExceptionForCat<ErrorCategory::NeedRetargettingError>& ex) {
                 const auto staleNs = [&] {
                     if (auto staleInfo = ex.extraInfo<StaleConfigInfo>()) {
                         return NamespaceString(staleInfo->getns());
@@ -383,32 +383,34 @@ void runCommand(OperationContext* opCtx,
                     throw;
                 }
 
-                if (!canRetry)
-                    throw;
-
-                LOG(2) << "Retrying command " << redact(request.body) << causedBy(ex);
-
+                // Send setShardVersion on this thread's versioned connections to shards (to support
+                // commands that use the legacy (ShardConnection) versioning protocol).
                 if (!MONGO_FAIL_POINT(doNotRefreshShardsOnRetargettingError)) {
                     ShardConnection::checkMyConnectionVersions(opCtx, staleNs.ns());
                 }
 
+                // Mark collection entry in cache as stale.
                 if (staleNs.isValid()) {
                     Grid::get(opCtx)->catalogCache()->invalidateShardedCollection(staleNs);
                 }
-
-                continue;
-            } catch (const ExceptionFor<ErrorCodes::StaleDbVersion>& e) {
-                Grid::get(opCtx)->catalogCache()->onStaleDatabaseVersion(e->getDb(),
-                                                                         e->getVersionReceived());
-                if (!canRetry)
-                    throw;
-                continue;
-            } catch (const ExceptionForCat<ErrorCategory::SnapshotError>& ex) {
-                LOG(2) << "Retrying command " << redact(request.body) << causedBy(ex);
-                if (!canRetry)
-                    throw;
+                if (canRetry) {
+                    continue;
+                }
+                throw;
+            } catch (const ExceptionFor<ErrorCodes::StaleDbVersion>& ex) {
+                // Mark database entry in cache as stale.
+                Grid::get(opCtx)->catalogCache()->onStaleDatabaseVersion(ex->getDb(),
+                                                                         ex->getVersionReceived());
+                if (canRetry) {
+                    continue;
+                }
+                throw;
+            } catch (const ExceptionForCat<ErrorCategory::SnapshotError>&) {
                 // Simple retry on any type of snapshot error.
-                continue;
+                if (canRetry) {
+                    continue;
+                }
+                throw;
             }
             MONGO_UNREACHABLE;
         }
@@ -766,12 +768,55 @@ void Strategy::explainFind(OperationContext* opCtx,
                                                            qr.getCollation());
             millisElapsed = timer.millis();
             break;
-        } catch (const ExceptionFor<ErrorCodes::StaleDbVersion>& e) {
-            Grid::get(opCtx)->catalogCache()->onStaleDatabaseVersion(e->getDb(),
-                                                                     e->getVersionReceived());
-            if (!canRetry) {
+        } catch (const ExceptionForCat<ErrorCategory::NeedRetargettingError>& ex) {
+            const auto staleNs = [&] {
+                if (auto staleInfo = ex.extraInfo<StaleConfigInfo>()) {
+                    return NamespaceString(staleInfo->getns());
+                } else if (auto implicitCreateInfo =
+                               ex.extraInfo<CannotImplicitlyCreateCollectionInfo>()) {
+                    return NamespaceString(implicitCreateInfo->getNss());
+                } else {
+                    throw;
+                }
+            }();
+
+            if (staleNs.isEmpty()) {
+                // This should be impossible but older versions tried incorrectly to handle
+                // it here.
+                log() << "Received a stale config error with an empty namespace while "
+                         "executing "
+                      << redact(explainCmd) << " : " << redact(ex);
                 throw;
             }
+
+            // Send setShardVersion on this thread's versioned connections to shards (to support
+            // commands that use the legacy (ShardConnection) versioning protocol).
+            if (!MONGO_FAIL_POINT(doNotRefreshShardsOnRetargettingError)) {
+                ShardConnection::checkMyConnectionVersions(opCtx, staleNs.ns());
+            }
+
+            // Mark collection entry in cache as stale.
+            if (staleNs.isValid()) {
+                Grid::get(opCtx)->catalogCache()->invalidateShardedCollection(staleNs);
+            }
+            if (canRetry) {
+                continue;
+            }
+            throw;
+        } catch (const ExceptionFor<ErrorCodes::StaleDbVersion>& ex) {
+            // Mark database entry in cache as stale.
+            Grid::get(opCtx)->catalogCache()->onStaleDatabaseVersion(ex->getDb(),
+                                                                     ex->getVersionReceived());
+            if (canRetry) {
+                continue;
+            }
+            throw;
+        } catch (const ExceptionForCat<ErrorCategory::SnapshotError>&) {
+            // Simple retry on any type of snapshot error.
+            if (canRetry) {
+                continue;
+            }
+            throw;
         }
     }