Github user gparai commented on a diff in the pull request:
https://github.com/apache/drill/pull/653#discussion_r88306202
--- Diff:
exec/java-exec/src/main/java/org/apache/drill/exec/store/parquet/Metadata.java
---
@@ -495,31 +499,65 @@ private ParquetFileMetadata_v3
getParquetFileMetadata_v3(ParquetTableMetadata_v3
* @param p
* @throws IOException
*/
- private void writeFile(ParquetTableMetadata_v3 parquetTableMetadata,
Path p) throws IOException {
+ private void writeFile(ParquetTableMetadata_v3 parquetTableMetadata,
String path) throws IOException {
JsonFactory jsonFactory = new JsonFactory();
jsonFactory.configure(Feature.AUTO_CLOSE_TARGET, false);
jsonFactory.configure(JsonParser.Feature.AUTO_CLOSE_SOURCE, false);
ObjectMapper mapper = new ObjectMapper(jsonFactory);
SimpleModule module = new SimpleModule();
module.addSerializer(ColumnMetadata_v3.class, new
ColumnMetadata_v3.Serializer());
mapper.registerModule(module);
- FSDataOutputStream os = fs.create(p);
+
+ // If multiple clients are updating metadata cache file concurrently,
the cache file
+ // can get corrupted. To prevent this, write to a unique temporary
file and then do
+ // atomic rename.
+ UUID randomUUID = UUID.randomUUID();
+ Path tmpPath = new Path(path, new String(METADATA_FILENAME + "." +
randomUUID));
+
+ FSDataOutputStream os = fs.create(tmpPath);
mapper.writerWithDefaultPrettyPrinter().writeValue(os,
parquetTableMetadata);
os.flush();
os.close();
+
+ // Use fileContext API as FileSystem rename is deprecated.
+ FileContext fileContext = FileContext.getFileContext(tmpPath.toUri());
+ Path finalPath = new Path(path, METADATA_FILENAME);
+
+ try {
+ fileContext.rename(tmpPath, finalPath, Options.Rename.OVERWRITE);
+ } catch (Exception e) {
+ logger.info("Rename from {} to {} failed", tmpPath.toString(),
finalPath.toString(), e);
+ }
--- End diff --
Should we add a `finally` section which removes the temp file if it exists
i.e. case rename failed?
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastruct...@apache.org or file a JIRA ticket
with INFRA.
---
