UnderReplicatedBlocks是HDFS中关于块复制的一个重要数据结构。在HDFS的高性能、高容错性体系中,总有一些原因促使HDFS系统内进行块复制工作,比如基于高性能的负载均衡、基于容错性的数据块副本数恢复等。普遍的,任何工作都会有一个优先级的问题,特别是这里的数据块复制,不可能简单的按照先入先出或者其他简单策略,比方说,基于容错性的数据块副本数恢复,特别是数据块副本仅有一个的数据块副本数恢复,其优先级肯定要比基于高性能的负载均衡高,所以数据块复制要有个优先级的概念,那么,数据块复制的优先级怎么确定,怎么存储?一切答案均在UnderReplicatedBlocks中,本文我们将开始分析UnderReplicatedBlocks。
UnderReplicatedBlocks专门用于存储待优先级的需要复制的数据块。何谓数据块复制优先级,我们看下UnderReplicatedBlocks类中的几个静态成员变量及其说明就会得到答案,如下:
/** The queue with the highest priority: {@value} */ // 最高优先级队列的优先级值0 static final int QUEUE_HIGHEST_PRIORITY = 0; /** The queue for blocks that are way below their expected value : {@value} */ // 第二优先级队列的优先级值1:主要针对低于副本数要求很多的数据块 static final int QUEUE_VERY_UNDER_REPLICATED = 1; /** The queue for "normally" under-replicated blocks: {@value} */ // 第三优先级队列的优先级值2:主要针对低于副本数要求不是很多,即一般情况的数据块 static final int QUEUE_UNDER_REPLICATED = 2; /** The queue for blocks that have the right number of replicas, * but which the block manager felt were badly distributed: {@value} */ // 第四优先级队列的优先级值3:主要针对副本数满足要求,但是数据块管理器BlockManager感觉严重分布不均 static final int QUEUE_REPLICAS_BADLY_DISTRIBUTED = 3; /** The queue for corrupt blocks: {@value} */ // 第五优先级队列的优先级值4:主要针对损坏的数据块 static final int QUEUE_WITH_CORRUPT_BLOCKS = 4;数据块复制优先级共分为五种级别,从高到底依次如下:
1、QUEUE_HIGHEST_PRIORITY = 0:最高优先级
主要针对数据块副本数非常的、严重的不足的情况,当前副本数低于期望值,且仅有1个或者干脆没有,比如副本数仅有1个,或者副本数干脆为0,但是还存在退役副本,这种情况最危险,数据最容易丢失,所以复制的优先级也最高;
2、QUEUE_VERY_UNDER_REPLICATED = 1:第二优先级
主要针对数据块副本数比较不足的情况,比上面的情况好点,当前副本数低于期望值,但是副本数大于1,其判断公式为当前副本数curReplicas乘以3还小于期望副本数expectedReplicas,这种情况也比较危险,数据也容易丢失,所以复制的优先级也很高;
3、QUEUE_UNDER_REPLICATED = 2:第三优先级
主要针对数据块副本数低于期望值,但是还不是很严重、很危急的情况;
4、QUEUE_REPLICAS_BADLY_DISTRIBUTED = 3:第四优先级
主要针对数据块已经有足够的副本数,但是没有足够的机架的情况,这是负载均衡等策略需要的产物;
5、QUEUE_WITH_CORRUPT_BLOCKS = 4:第五优先级
主要针对损坏的数据块的情况,其副本数位0,但是还没有退役副本,所以优先级最低,话说,这种数据块还需要哦复制吗?留个小小的疑问吧!
通过上面的说明,我们可以简单总结下:
当前副本数低于期望值时,如果当前副本数为1,甚至存在退役副本的情况下为0时,其复制优先级最高,如果当前副本数为0且没有退役副本,则复制优先级最低;如果当前副本数大于1,但是乘以3还小于期望副本数,处于比较危险的情况,则优先级次之,否则是第三优先级。而当当前副本数等于或高于期望值时,则可能是没有足够机架的情况,此时的优先级比最低优先级稍高,为第四优先级。
UnderReplicatedBlocks也提供了根据数据块及其副本情况来获取复制优先级的getPriority()方法,代码如下:
/** Return the priority of a block * 计算指定数据块复制优先级 * * @param block a under replicated block * @param curReplicas current number of replicas of the block * @param expectedReplicas expected number of replicas of the block * @return the priority for the blocks, between 0 and ({@link #LEVEL}-1) */ private int getPriority(Block block, int curReplicas, int decommissionedReplicas, int expectedReplicas) { // 参数校验:当前副本数curReplicas应大于等于0 assert curReplicas >= 0 : "Negative replicas!"; // 如果当前副本数curReplicas大于等于期望的副本数,则返回第四优先级队列的优先级值3 if (curReplicas >= expectedReplicas) { // 数据块已经有足够的副本数,但是没有足够的机架 // Block has enough copies, but not enough racks return QUEUE_REPLICAS_BADLY_DISTRIBUTED; } else if (curReplicas == 0) {// 如果当前副本数curReplicas为0 // If there are zero non-decommissioned replicas but there are // some decommissioned replicas, then assign them highest priority // 如果decommissionedReplicas大于0,返回最高优先级队列的优先级值0 // 即没有非退役副本,但是有一些退役的副本,那么我们需要分配给它们最高优先级 if (decommissionedReplicas > 0) { return QUEUE_HIGHEST_PRIORITY; } // all we have are corrupt blocks // 没有非退役副本,也没有退役副本,我们就认为它是个损坏的数据块,复制优先级最低,为第五优先级队列的优先级值4 return QUEUE_WITH_CORRUPT_BLOCKS; } else if (curReplicas == 1) {// 如果当前副本数curReplicas为1 //only on replica -risk of loss // highest priority // 仅仅有一个副本,有丢失的风险,所以赋予最高优先级0 return QUEUE_HIGHEST_PRIORITY; } else if ((curReplicas * 3) < expectedReplicas) { //there is less than a third as many blocks as requested; //this is considered very under-replicated // 如果如果当前副本数curReplicas乘以3还小于期望副本数expectedReplicas,返回第二优先级队列的优先级值1 return QUEUE_VERY_UNDER_REPLICATED; } else { //add to the normal queue for under replicated blocks // 一般的低于副本数的优先级,返回第三优先级队列的优先级值2 return QUEUE_UNDER_REPLICATED; }思路和上述介绍一样,大体逻辑如下:
1、如果当前副本数curReplicas大于等于期望的副本数,则返回第四优先级队列的优先级值3--QUEUE_REPLICAS_BADLY_DISTRIBUTED;
2、如果当前副本数curReplicas为0,且如果decommissionedReplicas大于0,返回最高优先级队列的优先级值0,没有非退役副本,也没有退役副本,我们就认为它是个损坏的数据块,复制优先级最低,为第五优先级队列的优先级值4;
3、如果当前副本数curReplicas为1,仅仅有一个副本,有丢失的风险,所以赋予最高优先级0;
4、如果如果当前副本数curReplicas乘以3还小于期望副本数expectedReplicas,返回第二优先级队列的优先级值1;
5、一般的低于副本数的优先级,返回第三优先级队列的优先级值2。
UnderReplicatedBlocks还提供了涉及复制优先级队列的成员变量,如下:
/** The total number of queues : {@value} */ // 队列总数 static final int LEVEL = 5; /** the queues themselves */ // 队列集合 private final List<LightWeightLinkedSet<Block>> priorityQueues = new ArrayList<LightWeightLinkedSet<Block>>(); /** Stores the replication index for each priority */ // 存储每个优先级对应的复制索引 private Map<Integer, Integer> priorityToReplIdx = new HashMap<Integer, Integer>(LEVEL);总的队列数目为5,而存储待复制不同优先级块集合的是priorityQueues列表,它是数据块集合LightWeightLinkedSet的列表,并且还提供了存储每种优先级对应的块复制索引的集合priorityToReplIdx,它是数字形式优先级priority到块在集合LightWeightLinkedSet中位置索引index的映射。
UnderReplicatedBlocks的构造函数如下:
/** Create an object. */ // 构造函数,创建一个对象 UnderReplicatedBlocks() { // 5个LightWeightLinkedSet集合,存储到priorityQueues列表中, // 并将优先级与复制索引的映射存储到priorityToReplIdx中 for (int i = 0; i < LEVEL; i++) { priorityQueues.add(new LightWeightLinkedSet<Block>()); priorityToReplIdx.put(i, 0); } }上来先构造5个LightWeightLinkedSet集合,并按照优先级由高到低的顺序,添加到列表priorityQueues中,并初始化每种块复制优先级对应的位置索引为0。
UnderReplicatedBlocks还提供了相应的添加、移除数据块及更新优先级方法,分别介绍如下:
1、添加数据块add()
/** add a block to a under replication queue according to its priority * @param block a under replication block * @param curReplicas current number of replicas of the block * @param decomissionedReplicas the number of decommissioned replicas * @param expectedReplicas expected number of replicas of the block * @return true if the block was added to a queue. */ synchronized boolean add(Block block, int curReplicas, int decomissionedReplicas, int expectedReplicas) { assert curReplicas >= 0 : "Negative replicas!"; // 根据入参数据块及其副本情况计算块复制的优先级priLevel int priLevel = getPriority(block, curReplicas, decomissionedReplicas, expectedReplicas); // 如果块复制优先级priLevel小于5(即是一个正确有效的优先级),并且 // 根据优先级priLevel从priorityQueues中取出相应块集合并将块添加入集合成功的话,返回true if(priLevel != LEVEL && priorityQueues.get(priLevel).add(block)) { if(NameNode.blockStateChangeLog.isDebugEnabled()) { NameNode.blockStateChangeLog.debug( "BLOCK* NameSystem.UnderReplicationBlock.add:" + block + " has only " + curReplicas + " replicas and need " + expectedReplicas + " replicas so is added to neededReplications" + " at priority level " + priLevel); } return true; } // 否则返回false return false; }添加数据块的add()方法比较简单,首先根据入参数据块及其副本情况调用getPriority()方法计算块复制的优先级priLevel,然后如果块复制优先级priLevel小于5(即是一个正确有效的优先级),并且根据优先级priLevel从priorityQueues中取出相应块集合并将块添加入集合成功的话,返回true,表示添加成功,否则返回false,表示添加失败。
2、移除数据块remove()
/** remove a block from a under replication queue */ synchronized boolean remove(Block block, int oldReplicas, int decommissionedReplicas, int oldExpectedReplicas) { // 先根据入参数据块及其副本情况,调用getPriority()方法计算块复制优先级priLevel int priLevel = getPriority(block, oldReplicas, decommissionedReplicas, oldExpectedReplicas); // 调用两个参数的remove()方法,移除数据块 return remove(block, priLevel); }
/** * Remove a block from the under replication queues. * * The priLevel parameter is a hint of which queue to query * first: if negative or >= {@link #LEVEL} this shortcutting * is not attmpted. * * If the block is not found in the nominated queue, an attempt is made to * remove it from all queues. * * <i>Warning:</i> This is not a synchronized method. * @param block block to remove * @param priLevel expected privilege level * @return true if the block was found and removed from one of the priority queues */ boolean remove(Block block, int priLevel) { // 如果优先级priLevel是正确有效的,且根据优先级priLevel从列表priorityQueues中 // 取出数据块集合后,从中移除数据块成功的话,返回true,表示移除成功 if(priLevel >= 0 && priLevel < LEVEL && priorityQueues.get(priLevel).remove(block)) { if(NameNode.blockStateChangeLog.isDebugEnabled()) { NameNode.blockStateChangeLog.debug( "BLOCK* NameSystem.UnderReplicationBlock.remove: " + "Removing block " + block + " from priority queue "+ priLevel); } return true; } else { // 否则,在给定优先级对应数据块集合中移除失败的话,尝试从所有优先级各自对应的队列中移除数据块, // 任何一个移除成功,均返回true,表示移除成功 // Try to remove the block from all queues if the block was // not found in the queue for the given priority level. for (int i = 0; i < LEVEL; i++) { if (priorityQueues.get(i).remove(block)) { if(NameNode.blockStateChangeLog.isDebugEnabled()) { NameNode.blockStateChangeLog.debug( "BLOCK* NameSystem.UnderReplicationBlock.remove: " + "Removing block " + block + " from priority queue "+ i); } return true; } } } // 最后,如果还不行的话,则返回false,表示移除失败 return false; }首先,在四个参数的remove()方法中,先根据入参数据块及其副本情况,调用getPriority()方法计算块复制优先级priLevel,然后调用两个参数的remove()方法,移除数据块;
其次,在两个参数的remove()方法中,如果优先级priLevel是正确有效的,且根据优先级priLevel从列表priorityQueues中取出数据块集合后,从中移除数据块成功的话,返回true,表示移除成功;否则,在给定优先级对应数据块集合中移除失败的话,尝试从所有优先级各自对应的队列中移除数据块,任何一个移除成功,均返回true,表示移除成功;最后,如果还不行的话,则返回false,表示移除失败。
3、更新优先级update()
/** * Recalculate and potentially update the priority level of a block. * * If the block priority has changed from before an attempt is made to * remove it from the block queue. Regardless of whether or not the block * is in the block queue of (recalculate) priority, an attempt is made * to add it to that queue. This ensures that the block will be * in its expected priority queue (and only that queue) by the end of the * method call. * @param block a under replicated block * @param curReplicas current number of replicas of the block * @param decommissionedReplicas the number of decommissioned replicas * @param curExpectedReplicas expected number of replicas of the block * @param curReplicasDelta the change in the replicate count from before * @param expectedReplicasDelta the change in the expected replica count from before */ synchronized void update(Block block, int curReplicas, int decommissionedReplicas, int curExpectedReplicas, int curReplicasDelta, int expectedReplicasDelta) { // curReplicas代表当前副本数,curReplicasDelta代表之前发生的副本数变化 // curExpectedReplicas代表当前期望副本数,expectedReplicasDelta代表之前发生的期望副本数变化 // 计算之前的副本数oldReplicas和之前的期望副本数oldExpectedReplicas int oldReplicas = curReplicas-curReplicasDelta; int oldExpectedReplicas = curExpectedReplicas-expectedReplicasDelta; // 计算当前的块复制优先级curPri int curPri = getPriority(block, curReplicas, decommissionedReplicas, curExpectedReplicas); // 计算之前的块复制优先级oldPri int oldPri = getPriority(block, oldReplicas, decommissionedReplicas, oldExpectedReplicas); if(NameNode.stateChangeLog.isDebugEnabled()) { NameNode.stateChangeLog.debug("UnderReplicationBlocks.update " + block + " curReplicas " + curReplicas + " curExpectedReplicas " + curExpectedReplicas + " oldReplicas " + oldReplicas + " oldExpectedReplicas " + oldExpectedReplicas + " curPri " + curPri + " oldPri " + oldPri); } // 如果之前优先级oldPri合法且不等于当前优先级curPri if(oldPri != LEVEL && oldPri != curPri) { // 调用remove()方法移除数据块 remove(block, oldPri); } // 如果当前优先级curPri合法,通过当前优先级curPri从priorityQueues列表中获取对应数据块集合并将数据块添加进去 if(curPri != LEVEL && priorityQueues.get(curPri).add(block)) { if(NameNode.blockStateChangeLog.isDebugEnabled()) { NameNode.blockStateChangeLog.debug( "BLOCK* NameSystem.UnderReplicationBlock.update:" + block + " has only "+ curReplicas + " replicas and needs " + curExpectedReplicas + " replicas so is added to neededReplications" + " at priority level " + curPri); } } }更新优先级update()方法用于当数据块副本数或期望副本数等发生变化时,调整数据块复制优先级,并调整其在UnderReplicatedBlocks中的相应存储位置。大体逻辑如下:
1、首先搞清楚几个参数:curReplicas代表当前副本数,curReplicasDelta代表之前发生的副本数变化,curExpectedReplicas代表当前期望副本数,expectedReplicasDelta代表之前发生的期望副本数变化;
2、计算之前的副本数oldReplicas和之前的期望副本数oldExpectedReplicas;
3、计算当前的块复制优先级curPri;
4、计算之前的块复制优先级oldPri;
5、如果之前优先级oldPri合法且不等于当前优先级curPri:调用remove()方法移除数据块;
6、如果当前优先级curPri合法,通过当前优先级curPri从priorityQueues列表中获取对应数据块集合并将数据块添加进去。
未完待续,更多精彩尽在《HDFS源码分析之UnderReplicatedBlocks(二)》!