wombatu-kun commented on code in PR #11559: URL: https://github.com/apache/hudi/pull/11559#discussion_r1682719482
########## rfc/rfc-80/rfc-80.md: ########## @@ -0,0 +1,158 @@ +<!-- + Licensed to the Apache Software Foundation (ASF) under one or more + contributor license agreements. See the NOTICE file distributed with + this work for additional information regarding copyright ownership. + The ASF licenses this file to You under the Apache License, Version 2.0 + (the "License"); you may not use this file except in compliance with + the License. You may obtain a copy of the License at + + http://www.apache.org/licenses/LICENSE-2.0 + + Unless required by applicable law or agreed to in writing, software + distributed under the License is distributed on an "AS IS" BASIS, + WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + See the License for the specific language governing permissions and + limitations under the License. +--> +# RFC-80: Support column families for wide tbles + +## Proposers + +- @xiarixiaoyao +- @wombatu-kun + +## Approvers + - + - + +## Status + +JIRA: https://issues.apache.org/jira/browse/HUDI- + +## Abstract + +In streaming processing, there are often scenarios where the table is widened. The current mainstream real-time stretching is completed through Flink's multi-layer join; +Flink's join will cache a large amount of data in the state backend. As the data set increases, the pressure on the Flink task state backend will gradually increase, and may even become unavailable. +In multi-layer join scenarios, this problem is more obvious. + +## Background +Currently, Hudi organizes data according to fileGroup granularity. The fileGroup is further divided into column clusters to introduce the columnFamily concept. +The organizational form of Hudi files is divided according to the following rules: +The data in the partition is divided into buckets according to hash; the files in each bucket are divided according to columnFamily; multiple colFamily files in the bucket form a completed fileGroup; when there is only one columnFamily, it degenerates into the native Hudi bucket table. + + + +After splitting the fileGroup by columnFamily, the naming rules for base files and log files change. We add the cfName suffix to all file names to facilitate Hudi itself to distinguish column families. The addition of this suffix is compatible with Hudi's original naming method and has no conflict. + + + +## Implementation + +### Constraints and Restrictions +1. The overall design relies on the non-blocking concurrent writing feature of Hudi 1.0. +2. Lower version Hudi cannot read and write column family tables. +3. Only MOR bucketed tables support setting column families. +4. Column families do not support repartitioning and renaming. +5. Schema evolution does not take effect on the current column family table. +6. Like native bucket tables, clustering operations are not supported. + +### Model change +After the column family is introduced, the storage structure of the entire Hudi bucket table changes: + + Review Comment: translated -- 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. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
