Merge pull request #4330 from atlanhq/APP-5713

APP-5713 Added new Error code for service unavailable
atlanhq · Mar 4, 2025 · 7c40f76 · 7c40f76
2 parents 89e4587 + 57618f7
commit 7c40f76
Show file tree

Hide file tree

Showing 2 changed files with 31 additions and 0 deletions.
diff --git a/...anus/src/main/java/org/apache/atlas/repository/graphdb/janus/AtlasElasticsearchQuery.java b/...anus/src/main/java/org/apache/atlas/repository/graphdb/janus/AtlasElasticsearchQuery.java
@@ -54,6 +54,8 @@
 import org.slf4j.LoggerFactory;
 
 import java.io.IOException;
+import java.net.SocketTimeoutException;
+import java.net.UnknownHostException;
 import java.util.*;
 import java.util.concurrent.CompletableFuture;
 import java.util.concurrent.Future;
@@ -180,6 +182,9 @@ private DirectIndexQueryResult runQueryWithLowLevelClient(SearchParams searchPar
             }
         } catch (IOException e) {
             LOG.error("Failed to execute direct query on ES {}", e.getMessage());
+
+            handleNetworkErrors(e);
+
             throw new AtlasBaseException(AtlasErrorCode.INDEX_SEARCH_FAILED, e.getMessage());
         }
     }
@@ -207,6 +212,9 @@ private Map<String, Object> runQueryWithLowLevelClient(String query) throws Atla
 
         } catch (IOException e) {
             LOG.error("Failed to execute direct query on ES {}", e.getMessage());
+
+            handleNetworkErrors(e);
+
             throw new AtlasBaseException(AtlasErrorCode.INDEX_SEARCH_FAILED, e.getMessage());
         }
     }
@@ -220,6 +228,9 @@ private Map<String, LinkedHashMap> runUpdateByQueryWithLowLevelClient(String que
 
         } catch (IOException e) {
             LOG.error("Failed to execute direct query on ES {}", e.getMessage());
+
+            handleNetworkErrors(e);
+
             throw new AtlasBaseException(AtlasErrorCode.INDEX_SEARCH_FAILED, e.getMessage());
         }
     }
@@ -271,6 +282,11 @@ private DirectIndexQueryResult performAsyncDirectIndexQuery(SearchParams searchP
             }
         }catch (Exception e) {
             LOG.error("Failed to execute direct query on ES {}", e.getMessage());
+
+            if (e instanceof IOException) {
+                handleNetworkErrors((IOException) e);
+            }
+
             throw new AtlasBaseException(AtlasErrorCode.INDEX_SEARCH_FAILED, e.getMessage());
         } finally {
             if (contextIdExists) {
@@ -286,6 +302,20 @@ private DirectIndexQueryResult performAsyncDirectIndexQuery(SearchParams searchP
         return result;
     }
 
+    /*
+     * Checks if the exception is a network-related issue and throws a SERVICE_UNAVAILABLE error.
+     */
+    private void handleNetworkErrors(Exception e) throws AtlasBaseException {
+        if (e instanceof SocketTimeoutException || e instanceof UnknownHostException ||
+                (e.getMessage() != null &&
+                        (e.getMessage().contains("Connection reset by peer") ||
+                                e.getMessage().contains("Network error") ||
+                                e.getMessage().contains("Connection refused")))) {
+            throw new AtlasBaseException(AtlasErrorCode.SERVICE_UNAVAILABLE,
+                    "Service is unavailable or a network error occurred: Elasticsearch - " + e.getMessage());
+        }
+    }
+
     /*
         * Process the request with the same search context ID and sequence number
         * @param searchParams

diff --git a/intg/src/main/java/org/apache/atlas/AtlasErrorCode.java b/intg/src/main/java/org/apache/atlas/AtlasErrorCode.java
@@ -279,6 +279,7 @@ public enum AtlasErrorCode {
     KEYCLOAK_INIT_FAILED(500, "ATLAS-500-00-022", "Failed to initialize keycloak client: {0}"),
 
     MAINTENANCE_MODE_ENABLED(503, "ATLAS-503-00-001", "Atlas is in maintenance mode for this specific operation. Please try again later."),
+    SERVICE_UNAVAILABLE(503, "ATLAS-503-00-002", "Service is unavailable or a network error occurred: {0}"),
 
     BATCH_SIZE_TOO_LARGE(406, "ATLAS-406-00-001", "Batch size is too large, please use a smaller batch size"),