yifan-c commented on code in PR #4399: URL: https://github.com/apache/cassandra/pull/4399#discussion_r2425436401
########## src/java/org/apache/cassandra/db/compression/ManualTrainingOptions.java: ########## @@ -0,0 +1,80 @@ +/* + * 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. + */ + +package org.apache.cassandra.db.compression; + +import java.util.Map; + +/** + * Configuration options for manual compression dictionary training. + * This class encapsulates the parsed and validated parameters needed for training. + */ +public class ManualTrainingOptions +{ + private final int maxSamplingDurationSeconds; + + public ManualTrainingOptions(int maxSamplingDurationSeconds) + { + if (maxSamplingDurationSeconds <= 0) + { + throw new IllegalArgumentException("maxSamplingDurationSeconds must be positive, got: " + maxSamplingDurationSeconds); + } + this.maxSamplingDurationSeconds = maxSamplingDurationSeconds; + } + + /** + * Parse options from a string map, typically from JMX/MBean calls. + * + * @param options the string map containing training options + * @return parsed and validated ManualTrainingOptions + * @throws IllegalArgumentException if required parameters are missing or invalid + */ + public static ManualTrainingOptions fromStringMap(Map<String, String> options) + { + if (options == null || !options.containsKey("maxSamplingDurationSeconds")) + { + throw new IllegalArgumentException("maxSamplingDurationSeconds parameter is required for manual dictionary training"); Review Comment: The typical reason for training failure (after about 10 minutes) is that no `flush` is invoked with the `traincompressiondictionary` command. Otherwise, within 10 minutes, there should normally be enough samples collected. To clarify the current implementation: the termination condition for training is already defined as either reaching the maximum number of samples or the maximum sampling duration. The maximum sample size (`compression_dictionary_training_max_total_sample_size`) is configured in `cassandra.yaml`. The trainer starts the dictionary training once either of these thresholds is reached. I agree that we should avoid shipping foot-guns. Based on the feedback, I see problems with the write-based sampling approach. It is too complex since it requires running another command (`flush`) to ensure success. Frequent flushes can also lead to undesirable side effects, such as producing many small SSTables. On the other hand, SSTable-based sampling works better for manual training. It provides a more deterministic and self-contained process, and ensures the command terminates without having to worry about the sampling duration or manual intervention. In contrast, write-based sampling seems more suitable for automatic (auto-training) scenarios. Given these points, should we remove write-based sampling from the manual `traincompressiondictionary` command? What do you think? -- 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] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
