/ 中存储网

hadoop容灾能力测试实验

2015-08-10 12:06:30 来源:中存储网

本次hadoop容灾能力测试实验简单来讲就是

1. put 一个600M文件,分散3个replica x, 9个block 共18个blocks到4个datanode

2. 我关掉了两个datanode,使得大部分的block只在一个datanode上存在,但因为9个很分散,所以文件能正确取回(靠的是checksum来计算文件值)

3. hadoop namenode很迅速的复制了仅有一个replica的block使之成为 3 replica(2) but only found 2

4. 我再关掉一个datanode,结果发现每个datanode被很均衡的分配了block,这样即使只有一个datanode,也因为之前有确保2个replicas的比率,所以依然healthy

5. 我从这个仅存的datanode中删除一个blk,namenode report这个文件corrupt,(我其实一直很希望能进safemode,结果-safemode get一直是OFF)

6. 然后我启动另外一个datanode,30秒不到,这个missing的block被从这个新启动的datanode中迅速“扩展”为2个replicas

容灾性非常可靠,如果使用至少三个rack的话,数据会非常坚挺,对HADOOP信任值 level up!

首先来了解一下HDFS的一些基本特性

HDFS设计基础与目标

硬件错误是常态。因此需要冗余
流式数据访问。即数据批量读取而非随机读写,Hadoop擅长做的是数据分析而不是事务处理
大规模数据集
简单一致性模型。为了降低系统复杂度,对文件采用一次性写多次读的逻辑设计,即是文件一经写入,关闭,就再也不能修改
程序采用“数据就近”原则分配节点执行
HDFS体系结构

NameNode
DataNode
事务日志
映像文件
SecondaryNameNode
Namenode

管理文件系统的命名空间
记录每个文件数据块在各个Datanode上的位置和副本信息
协调客户端对文件的访问
记录命名空间内的改动或空间本身属性的改动
Namenode使用事务日志记录HDFS元数据的变化。使用映像文件存储文件系统的命名空间,包括文件映射,文件属性等
Datanode

负责所在物理节点的存储管理
一次写入,多次读取(不修改)
文件由数据块组成,典型的块大小是64MB
数据块尽量散布道各个节点
读取数据流程

客户端要访问HDFS中的一个文件
首先从namenode获得组成这个文件的数据块位置列表
根据列表知道存储数据块的datanode
访问datanode获取数据
Namenode并不参与数据实际传输
HDFS的可靠性

冗余副本策略
机架策略
心跳机制
安全模式
使用文件块的校验和 Checksum来检查文件的完整性
回收站
元数据保护
快照机制
我分别试验了冗余副本策略/心跳机制/安全模式/回收站。下面实验是关于冗余副本策略的。

环境:

Namenode/Master/jobtracker: h1/192.168.221.130
SecondaryNameNode: h1s/192.168.221.131
四个Datanode: h2~h4 (IP段:142~144)
为以防文件太小只有一个文件块(block/blk),我们准备一个稍微大一点的(600M)的文件,使之能分散分布到几个datanode,再停掉其中一个看有没有问题。 
先来put一个文件(为了方便起见,建议将hadoop/bin追加到$Path变量后 
:hadoop fs –put ~/Documents/IMMAUSWX201304 
结束后,我们想查看一下文件块的情况,可以去网页上看,也可以在namenode上使用fsck命令来检查一下,关于fsck命令
:bin/hadoop fsck /user/hadoop_admin/in/bigfile  -files -blocks -locations < ~/hadoopfiles/log1.txt 
下面打印结果说明 个600M文件被划分为9个64M的blocks,并且被分散到我当前所有datanode上(共4个),看起来比较平均, 

/user/hadoop_admin/in/bigfile/USWX201304 597639882 bytes, 9 block(s):  OK 
0. blk_-4541681964616523124_1011 len=67108864 repl=3 [192.168.221.131:50010, 192.168.221.142:50010, 192.168.221.144:50010] 
1. blk_4347039731705448097_1011 len=67108864 repl=3 [192.168.221.143:50010, 192.168.221.131:50010, 192.168.221.144:50010] 
2. blk_-4962604929782655181_1011 len=67108864 repl=3 [192.168.221.142:50010, 192.168.221.143:50010, 192.168.221.144:50010] 
3. blk_2055128947154747381_1011 len=67108864 repl=3 [192.168.221.143:50010, 192.168.221.142:50010, 192.168.221.144:50010] 
4. blk_-2280734543774885595_1011 len=67108864 repl=3 [192.168.221.131:50010, 192.168.221.142:50010, 192.168.221.144:50010] 
5. blk_6802612391555920071_1011 len=67108864 repl=3 [192.168.221.143:50010, 192.168.221.142:50010, 192.168.221.144:50010] 
6. blk_1890624110923458654_1011 len=67108864 repl=3 [192.168.221.143:50010, 192.168.221.142:50010, 192.168.221.144:50010] 
7. blk_226084029380457017_1011 len=67108864 repl=3 [192.168.221.143:50010, 192.168.221.131:50010, 192.168.221.144:50010] 
8. blk_-1230960090596945446_1011 len=60768970 repl=3 [192.168.221.142:50010, 192.168.221.143:50010, 192.168.221.144:50010]

Status: HEALTHY 
Total size:    597639882 B 
Total dirs:    0 
Total files:   1 
Total blocks (validated):      9 (avg. block size 66404431 B) 
Minimally replicated blocks:   9 (100.0 %) 
Over-replicated blocks:        0 (0.0 %) 
Under-replicated blocks:       0 (0.0 %) 
Mis-replicated blocks:         0 (0.0 %) 
Default replication factor:    3 
Average block replication:     3.0 
Corrupt blocks:                0 
Missing replicas:              0 (0.0 %) 
Number of data-nodes:          4 
Number of racks:               1

h1s,h2,h3,h4四个DD全部参与,跑去h2 (142),h3(143) stop datanode, 从h4上面get,发现居然能够get回,而且初步来看,size正确,看一下上图中黄底和绿底都DEAD了,每个blk都有源可以取回,所以GET后数据仍然是完整的,从这点看hadoop确实是强大啊,load balancing也做得很不错,数据看上去很坚强,容错性做得不错

1
再检查一下,我本来想测试safemode的,结果隔一会一刷,本来有几个blk只有1个livenode的,现在又被全部复制为确保每个有2个了!    

hadoop_admin@h1:~/hadoop-0.20.2$ hadoop fsck /user/hadoop_admin/in/bigfile  -files -blocks -locations 
/user/hadoop_admin/in/bigfile/USWX201304 597639882 bytes, 9 block(s):  
Under replicated blk_-4541681964616523124_1011. Target Replicas is 3 but found 2 replica(s). 
Under replicated blk_4347039731705448097_1011. Target Replicas is 3 but found 2 replica(s). 
Under replicated blk_-4962604929782655181_1011. Target Replicas is 3 but found 2 replica(s). 
Under replicated blk_2055128947154747381_1011. Target Replicas is 3 but found 2 replica(s). 
Under replicated blk_-2280734543774885595_1011. Target Replicas is 3 but found 2 replica(s). 
Under replicated blk_6802612391555920071_1011. Target Replicas is 3 but found 2 replica(s). 
Under replicated blk_1890624110923458654_1011. Target Replicas is 3 but found 2 replica(s). 
Under replicated blk_226084029380457017_1011. Target Replicas is 3 but found 2 replica(s). 
Under replicated blk_-1230960090596945446_1011. Target Replicas is 3 but found 2 replica(s). 
0. blk_-4541681964616523124_1011 len=67108864 repl=2 [192.168.221.131:50010, 192.168.221.144:50010] 
1. blk_4347039731705448097_1011 len=67108864 repl=2 [192.168.221.144:50010, 192.168.221.131:50010]
2. blk_-4962604929782655181_1011 len=67108864 repl=2 [192.168.221.144:50010, 192.168.221.131:50010] 
3. blk_2055128947154747381_1011 len=67108864 repl=2 [192.168.221.144:50010, 192.168.221.131:50010]
4. blk_-2280734543774885595_1011 len=67108864 repl=2 [192.168.221.131:50010, 192.168.221.144:50010] 
5. blk_6802612391555920071_1011 len=67108864 repl=2 [192.168.221.144:50010, 192.168.221.131:50010]
6. blk_1890624110923458654_1011 len=67108864 repl=2 [192.168.221.144:50010, 192.168.221.131:50010]
7. blk_226084029380457017_1011 len=67108864 repl=2 [192.168.221.144:50010, 192.168.221.131:50010] 
8. blk_-1230960090596945446_1011 len=60768970 repl=2 [192.168.221.144:50010, 192.168.221.131:50010]

我决定再关一个datanode,结果等了好半天也没见namenode发现它死了,这是因为心跳机制,datanode每隔3秒会向namenode发送heartbeat指令表明它的存活,但如果namenode很长时间(5~10分钟看设置)没有收到heartbeat即认为这个NODE死掉了,就会做出BLOCK的复制操作,以保证有足够的replica来保证数据有足够的容灾/错性,现在再打印看看,发现因为只有一个live datanode,所以现在每个blk都有且只有一份

hadoop_admin@h1:~$ hadoop fsck /user/hadoop_admin/in/bigfile -files -blocks -locations 
/user/hadoop_admin/in/bigfile/USWX201304 597639882 bytes, 9 block(s):  Under replicated blk_-4541681964616523124_1011. Target Replicas is 3 but found 1 replica(s). 
Under replicated blk_4347039731705448097_1011. Target Replicas is 3 but found 1 replica(s). 
Under replicated blk_-4962604929782655181_1011. Target Replicas is 3 but found 1 replica(s). 
Under replicated blk_2055128947154747381_1011. Target Replicas is 3 but found 1 replica(s). 
Under replicated blk_-2280734543774885595_1011. Target Replicas is 3 but found 1 replica(s). 
Under replicated blk_6802612391555920071_1011. Target Replicas is 3 but found 1 replica(s). 
Under replicated blk_1890624110923458654_1011. Target Replicas is 3 but found 1 replica(s). 
Under replicated blk_226084029380457017_1011. Target Replicas is 3 but found 1 replica(s). 
Under replicated blk_-1230960090596945446_1011. Target Replicas is 3 but found 1 replica(s). 

我现在把其中一个BLK从这个仅存的Datanode中移走使之corrupt,我想实验,重启一个DATANODE后,会不会复员 
hadoop_admin@h4:/hadoop_run/data/current$ mv blk_4347039731705448097_1011* ~/Documents/ 
然后为了不必要等8分钟DN发block report,我手动修改了h4的dfs.blockreport.intervalMsec值为30000,stop datanode,再start (另外,你应该把hadoop/bin也加入到Path变量后面,这样你可以不带全路径执行hadoop命令,结果,检测它已被损坏 
hadoop_admin@h1:~$ hadoop fsck /user/hadoop_admin/in/bigfile -files -blocks -locations 

/user/hadoop_admin/in/bigfile/USWX201304 597639882 bytes, 9 block(s):  Under replicated blk_-4541681964616523124_1011. Target Replicas is 3 but found 1 replica(s).

/user/hadoop_admin/in/bigfile/USWX201304: CORRUPT block blk_4347039731705448097
Under replicated blk_-4962604929782655181_1011. Target Replicas is 3 but found 1 replica(s). 
Under replicated blk_2055128947154747381_1011. Target Replicas is 3 but found 1 replica(s). 
Under replicated blk_-2280734543774885595_1011. Target Replicas is 3 but found 1 replica(s). 
Under replicated blk_6802612391555920071_1011. Target Replicas is 3 but found 1 replica(s). 
Under replicated blk_1890624110923458654_1011. Target Replicas is 3 but found 1 replica(s). 
Under replicated blk_226084029380457017_1011. Target Replicas is 3 but found 1 replica(s). 
Under replicated blk_-1230960090596945446_1011. Target Replicas is 3 but found 1 replica(s). 
MISSING 1 blocks of total size 67108864 B
0. blk_-4541681964616523124_1011 len=67108864 repl=1 [192.168.221.144:50010] 
1. blk_4347039731705448097_1011 len=67108864 MISSING!
2. blk_-4962604929782655181_1011 len=67108864 repl=1 [192.168.221.144:50010] 
3. blk_2055128947154747381_1011 len=67108864 repl=1 [192.168.221.144:50010] 
4. blk_-2280734543774885595_1011 len=67108864 repl=1 [192.168.221.144:50010] 
5. blk_6802612391555920071_1011 len=67108864 repl=1 [192.168.221.144:50010] 
6. blk_1890624110923458654_1011 len=67108864 repl=1 [192.168.221.144:50010] 
7. blk_226084029380457017_1011 len=67108864 repl=1 [192.168.221.144:50010] 
8. blk_-1230960090596945446_1011 len=60768970 repl=1 [192.168.221.144:50010]

Status: CORRUPT
Total size:    597639882 B 
Total dirs:    0 
Total files:   1 
Total blocks (validated):      9 (avg. block size 66404431 B) 
   ******************************** 
   CORRUPT FILES:        1 
   MISSING BLOCKS:       1 
   MISSING SIZE:         67108864 B 
   CORRUPT BLOCKS:       1 
   ******************************** 
Minimally replicated blocks:   8 (88.888885 %) 
Over-replicated blocks:        0 (0.0 %) 
Under-replicated blocks:       8 (88.888885 %) 
Mis-replicated blocks:         0 (0.0 %) 
Default replication factor:    3 
Average block replication:     0.8888889 
Corrupt blocks:                1 
Missing replicas:              16 (200.0 %) 
Number of data-nodes:          1 
Number of racks:               1


The filesystem under path '/user/hadoop_admin/in/bigfile' is CORRUPT

我现在启动一个DATANODE h1s(131),结果很快的在30秒之内,它就被hadoop原地满HP复活了,现在每个blk都有了两份replica 
hadoop_admin@h1:~$ hadoop fsck /user/hadoop_admin/in/bigfile -files -blocks -locations 
/user/hadoop_admin/in/bigfile/USWX201304 597639882 bytes, 9 block(s):  Under replicated blk_-4541681964616523124_1011. Target Replicas is 3 but found 2 replica(s). 
Under replicated blk_4347039731705448097_1011. Target Replicas is 3 but found 2 replica(s). 
Under replicated blk_-4962604929782655181_1011. Target Replicas is 3 but found 2 replica(s). 
Under replicated blk_2055128947154747381_1011. Target Replicas is 3 but found 2 replica(s). 
Under replicated blk_-2280734543774885595_1011. Target Replicas is 3 but found 2 replica(s). 
Under replicated blk_6802612391555920071_1011. Target Replicas is 3 but found 2 replica(s). 
Under replicated blk_1890624110923458654_1011. Target Replicas is 3 but found 2 replica(s). 
Under replicated blk_226084029380457017_1011. Target Replicas is 3 but found 2 replica(s). 
Under replicated blk_-1230960090596945446_1011. Target Replicas is 3 but found 2 replica(s). 
0. blk_-4541681964616523124_1011 len=67108864 repl=2 [192.168.221.144:50010, 192.168.221.131:50010] 
1. blk_4347039731705448097_1011 len=67108864 repl=2 [192.168.221.131:50010, 192.168.221.144:50010]
2. blk_-4962604929782655181_1011 len=67108864 repl=2 [192.168.221.144:50010, 192.168.221.131:50010] 
3. blk_2055128947154747381_1011 len=67108864 repl=2 [192.168.221.144:50010, 192.168.221.131:50010]
4. blk_-2280734543774885595_1011 len=67108864 repl=2 [192.168.221.144:50010, 192.168.221.131:50010] 
5. blk_6802612391555920071_1011 len=67108864 repl=2 [192.168.221.144:50010, 192.168.221.131:50010]
6. blk_1890624110923458654_1011 len=67108864 repl=2 [192.168.221.144:50010, 192.168.221.131:50010]
7. blk_226084029380457017_1011 len=67108864 repl=2 [192.168.221.144:50010, 192.168.221.131:50010] 
8. blk_-1230960090596945446_1011 len=60768970 repl=2 [192.168.221.144:50010, 192.168.221.131:50010]

发现这个文件被从131成功复制回了144 (h4)。

结论:HADOOP容灾太坚挺了,我现在坚信不疑了!

另外有一个没有粘出来的提示就是,h4 datanode上有不少重新format遗留下来的badLinkBlock,在重新put同一个文件的时候,hadoop将那些老旧残留的block文件全部都删除了。这说明它是具有删除无效bad block的功能的。