shanemkm commented on a change in pull request #2246:
URL: https://github.com/apache/hadoop/pull/2246#discussion_r480718118
##########
File path:
hadoop-tools/hadoop-azure/src/main/java/org/apache/hadoop/fs/azurebfs/services/AbfsClient.java
##########
@@ -271,10 +272,67 @@ public AbfsRestOperation deleteFilesystem() throws
AzureBlobFileSystemException
return op;
}
- public AbfsRestOperation createPath(final String path, final boolean isFile,
final boolean overwrite,
- final String permission, final String
umask,
- final boolean isAppendBlob) throws
AzureBlobFileSystemException {
+ public AbfsRestOperation createPath(final String path,
+ final boolean isFile,
+ final boolean overwrite,
+ final String permission,
+ final String umask,
+ final boolean isAppendBlob) throws AzureBlobFileSystemException {
+ String operation = isFile
+ ? SASTokenProvider.CREATE_FILE_OPERATION
+ : SASTokenProvider.CREATE_DIRECTORY_OPERATION;
+
+ // HDFS FS defaults overwrite behaviour to true for create file which leads
+ // to majority create API traffic with overwrite=true. In some cases, this
+ // will end in race conditions at backend with parallel operations issued
to
+ // same path either by means of the customer workload or ABFS driver retry.
+ // Disabling the create overwrite default setting to false should
+ // significantly reduce the chances for such race conditions.
+ boolean isFirstAttemptToCreateWithoutOverwrite = false;
+ if (isFile && overwrite
+ && abfsConfiguration.isDefaultCreateOverwriteDisabled()) {
+ isFirstAttemptToCreateWithoutOverwrite = true;
+ }
+
+ AbfsRestOperation op = null;
+ // Query builder
+ final AbfsUriQueryBuilder abfsUriQueryBuilder =
createDefaultUriQueryBuilder();
+ abfsUriQueryBuilder.addQuery(QUERY_PARAM_RESOURCE,
+ operation.equals(SASTokenProvider.CREATE_FILE_OPERATION)
+ ? FILE
+ : DIRECTORY);
+ if (isAppendBlob) {
+ abfsUriQueryBuilder.addQuery(QUERY_PARAM_BLOBTYPE, APPEND_BLOB_TYPE);
+ }
+
+ appendSASTokenToQuery(path, operation, abfsUriQueryBuilder);
+
+ try {
+ op = createPathImpl(path, abfsUriQueryBuilder,
Review comment:
In createPathImpl, it will go through retries with backoff inside of
that, correct? Just want to make sure since our observation was that those
retries almost always succeed, hence why this change would remove the race
condition/stale write requests almost entirely
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: common-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: common-issues-h...@hadoop.apache.org
