Support multi-tenant RAM buffers for IndexWriter #13951
There are no files selected for viewing
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,198 @@ | ||
| /* | ||
| * 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.lucene.index; | ||
|
|
||
| import java.io.IOException; | ||
| import java.util.HashMap; | ||
| import java.util.Map; | ||
| import java.util.concurrent.atomic.AtomicInteger; | ||
| import java.util.concurrent.locks.ReentrantLock; | ||
|
|
||
| /** | ||
| * For managing multiple instances of {@link IndexWriter} sharing the same buffer (configured by | ||
| * {@link IndexWriterConfig#setRAMBufferSizeMB}) | ||
|
There was a problem hiding this comment. It should be the other way around in my opinion, the RAM buffer size should be on There was a problem hiding this comment. Sorry I'm probably missing something here - so I get what you're saying about having the There was a problem hiding this comment.
I think the idea is to be able to create IndexWriters that don't share their RAM buffer limit with other writers. Maybe we could just set There was a problem hiding this comment. Hmm I took a slightly different approach, will publish a new PR soon. Maybe we can discuss it there but pretty much I just do what @jpountz suggested and move the |
||
| */ | ||
| public class IndexWriterRAMManager { | ||
| private final LinkedIdToWriter idToWriter = new LinkedIdToWriter(); | ||
| private final AtomicInteger idGenerator = new AtomicInteger(); | ||
| private double ramBufferSizeMB; | ||
|
|
||
| /** | ||
| * Default constructor | ||
| * | ||
| * @param ramBufferSizeMB the RAM buffer size to use between all registered {@link IndexWriter} | ||
| * instances | ||
| */ | ||
| public IndexWriterRAMManager(double ramBufferSizeMB) { | ||
| if (ramBufferSizeMB != IndexWriterConfig.DISABLE_AUTO_FLUSH && ramBufferSizeMB <= 0.0) { | ||
| throw new IllegalArgumentException("ramBufferSize should be > 0.0 MB when enabled"); | ||
| } | ||
| this.ramBufferSizeMB = ramBufferSizeMB; | ||
| } | ||
|
|
||
| /** Set the buffer size for this manager */ | ||
| public void setRamBufferSizeMB(double ramBufferSizeMB) { | ||
| this.ramBufferSizeMB = ramBufferSizeMB; | ||
| } | ||
|
|
||
| /** Get the buffer size assigned to this manager */ | ||
| public double getRamBufferSizeMB() { | ||
| return ramBufferSizeMB; | ||
| } | ||
|
|
||
| /** | ||
| * Calls {@link IndexWriter#flushNextBuffer()} in a round-robin fashion starting from the first | ||
| * writer added that has not been removed yet. Subsequent calls will flush the next writer in line | ||
| * and eventually loop back to the beginning. Returns the flushed writer id for testing | ||
| */ | ||
| public int flushRoundRobin() throws IOException { | ||
| return idToWriter.flushRoundRobin(); | ||
| } | ||
|
|
||
| /** Gets the number of writers registered with this ram manager */ | ||
| public int getWriterCount() { | ||
| return idToWriter.size(); | ||
| } | ||
|
|
||
| /** Registers a writer can returns the associated ID */ | ||
| protected int registerWriter(IndexWriter writer) { | ||
| int id = idGenerator.incrementAndGet(); | ||
| idToWriter.addWriter(writer, id); | ||
| return id; | ||
| } | ||
|
|
||
| /** Removes a writer given the writer's ide, protected for testing */ | ||
| protected void removeWriter(int id) { | ||
| idToWriter.removeWriter(id); | ||
| } | ||
|
|
||
| /** | ||
| * Will call {@link IndexWriter#ramBytesUsed()} for the writer id passed in, and then updates the | ||
| * total ram using that value and returns it | ||
| */ | ||
| public long updateAndGetCurrentBytesUsed(int id) { | ||
| return idToWriter.getTotalRamTracker(id); | ||
| } | ||
|
|
||
| private static class LinkedIdToWriter { | ||
|
There was a problem hiding this comment. How about using a java Queue implementation instead of the custom linked-list logic? You could round-robin on elements by removing, processing and add them back to the queue. I suppose this queue size would be small, so array deque and linked lists are both fine? We can also get some thread safe implementations out of the box. There was a problem hiding this comment. Hmm, so I looked into using both |
||
| private final Map<Integer, IndexWriterNode> idToWriterNode = new HashMap<>(); | ||
| private IndexWriterNode first; | ||
| private IndexWriterNode last; | ||
| private long totalRamTracker; | ||
|
|
||
| private final ReentrantLock lock = new ReentrantLock(); | ||
|
|
||
| // for round-robin flushing | ||
| private int lastIdFlushed = -1; | ||
|
|
||
| void addWriter(IndexWriter writer, int id) { | ||
| synchronized (lock) { | ||
| IndexWriterNode node = new IndexWriterNode(writer, id); | ||
| if (idToWriterNode.isEmpty()) { | ||
| first = node; | ||
| last = node; | ||
| } | ||
| node.next = first; | ||
| last.next = node; | ||
| node.prev = last; | ||
| last = node; | ||
| first.prev = node; | ||
| idToWriterNode.put(id, node); | ||
| } | ||
| } | ||
|
|
||
| void removeWriter(int id) { | ||
| synchronized (lock) { | ||
| if (idToWriterNode.containsKey(id)) { | ||
| IndexWriterNode nodeToRemove = idToWriterNode.remove(id); | ||
| totalRamTracker -= nodeToRemove.ram; | ||
| if (idToWriterNode.isEmpty()) { | ||
| first = null; | ||
| last = null; | ||
| lastIdFlushed = -1; | ||
| return; | ||
| } | ||
| if (id == lastIdFlushed) { | ||
| lastIdFlushed = nodeToRemove.prev.id; | ||
| } | ||
| nodeToRemove.prev.next = nodeToRemove.next; | ||
| nodeToRemove.next.prev = nodeToRemove.prev; | ||
| if (nodeToRemove == first) { | ||
| first = nodeToRemove.next; | ||
| } | ||
| if (nodeToRemove == last) { | ||
| last = nodeToRemove.prev; | ||
| } | ||
| } | ||
| } | ||
| } | ||
|
|
||
| // Returns the writer id that we attempted to flush (for testing purposes) | ||
| int flushRoundRobin() throws IOException { | ||
| synchronized (lock) { | ||
| if (idToWriterNode.isEmpty()) { | ||
| return -1; | ||
| } | ||
| int idToFlush; | ||
| if (lastIdFlushed == -1) { | ||
| idToFlush = first.id; | ||
| } else { | ||
| idToFlush = idToWriterNode.get(lastIdFlushed).next.id; | ||
| } | ||
| idToWriterNode.get(idToFlush).writer.flushNextBuffer(); | ||
| lastIdFlushed = idToFlush; | ||
| return idToFlush; | ||
| } | ||
| } | ||
|
|
||
| long getTotalRamTracker(int id) { | ||
| synchronized (lock) { | ||
| if (idToWriterNode.isEmpty()) { | ||
| return 0; | ||
| } | ||
| if (idToWriterNode.containsKey(id) == false) { | ||
| return totalRamTracker; | ||
| } | ||
| long oldRAMBytesUsed = idToWriterNode.get(id).ram; | ||
| long newRAMBytesUsed = idToWriterNode.get(id).writer.ramBytesUsed(); | ||
| idToWriterNode.get(id).ram = newRAMBytesUsed; | ||
| totalRamTracker += newRAMBytesUsed - oldRAMBytesUsed; | ||
| return totalRamTracker; | ||
| } | ||
| } | ||
|
|
||
| int size() { | ||
| synchronized (lock) { | ||
| return idToWriterNode.size(); | ||
| } | ||
| } | ||
|
|
||
| private static class IndexWriterNode { | ||
| IndexWriter writer; | ||
| int id; | ||
| long ram; | ||
| IndexWriterNode next; | ||
| IndexWriterNode prev; | ||
|
|
||
| IndexWriterNode(IndexWriter writer, int id) { | ||
| this.writer = writer; | ||
| this.id = id; | ||
| this.ram = writer.ramBytesUsed(); | ||
| } | ||
| } | ||
| } | ||
| } | ||
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Any reason for not making this change in the default constructor? We could avoid making changes to all the tests.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I think the reason I had to make all the changes in the test was cause I added. this constructor:
public IndexWriterConfig(IndexWriterRAMManager indexWriterRAMManager) {And then in the tests there were a bunch of
new IndexWriterConfig(null)calls which became ambiguous. I think that this constructor is potentially useful which is why I took the hit and changed all those tests, but I can remove it to avoid all those test changes?