hdfs无效块修复

登录到有无效块的机器后操作。

hdfs fsck /       后面的路径是指定检查哪个目录下有无效块，根据丢失文件目录可以细化，以提高检查速度

该命令会返回

FSCK started by hdfs (auth:SIMPLE) from /xxx.xxx.xxx for path / at Fri Jul 26 14:37:29 CST 2019
....................................................................................................
....................................................................................................
....................................................................................................
....................................................................................................
................................................................................................
/.../c9ddcb18-51c0-4fa7-bdab-daa5079bc094/rdd-1061487/_partitioner: Under replicated BP-770033680-172.20.36.31-1562142528050:blk_1073943459_202678. Target Replicas is 3 but found 2 live replica(s), 0 decommissioned replica(s), 0 decommissioning replica(s).
.
/.../c9ddcb18-51c0-4fa7-bdab-daa5079bc094/rdd-1061487/part-00000: Under replicated BP-770033680-172.20.36.31-1562142528050:blk_1073943450_202669. Target Replicas is 3 but found 2 live replica(s), 0 decommissioned replica(s), 0 decommissioning replica(s).
.
/.../c9ddcb18-51c0-4fa7-bdab-daa5079bc094/rdd-1061487/part-00001: Under replicated BP-770033680-172.20.36.31-1562142528050:blk_1073943456_202675. Target Replicas is 3 but found 2 live replica(s), 0 decommissioned replica(s), 0 decommissioning replica(s).
.
/.../c9ddcb18-51c0-4fa7-bdab-daa5079bc094/rdd-1061487/part-00002: Under replicated BP-770033680-172.20.36.31-1562142528050:blk_1073943452_202671. Target Replicas is 3 but found 2 live replica(s), 0 decommissioned replica(s), 0 decommissioning replica(s).

. . .

. . . 

. . . 

Total size: 64294925737 B (Total open files size: 672396467 B)
Total dirs: 2743
Total files: 2439
Total symlinks: 0 (Files currently being written: 22)
Total blocks (validated): 2752 (avg. block size 23362981 B) (Total open file blocks (not validated): 26)
Minimally replicated blocks: 2752 (100.0 %)
Over-replicated blocks: 0 (0.0 %)
Under-replicated blocks: 22 (0.7994186 %)
Mis-replicated blocks: 0 (0.0 %)
Default replication factor: 3
Average block replication: 2.9920058
Corrupt blocks: 0
Missing replicas: 22 (0.26647288 %)
Number of data-nodes: 3
Number of racks: 1
FSCK ended at Fri Jul 26 14:37:29 CST 2019 in 48 milliseconds

根据打印的块名称，找到文件系统中存储的块文件和块meta信息，将快文件和meta信息删除掉。

hadoop debug recoverLease -path $filePath -retries 10  恢复filePath对应的文件，如：/.../c9ddcb18-51c0-4fa7-bdab-daa5079bc094/rdd-1061487/part-00000

由于恢复块太多，使用脚本批量执行

#!/bin/bash
file_name=$1
path=$2
if [ -z "${file_name}" ];then
  echo "Pls input file path"
    exit 1
fi

cat ${file_name}| while read line
do
  if [ -z $line ]; then
    continue
  fi
  hadoop debug recoverLease -path $line -retries 10
done

完成恢复后需要重启该节点后，cdh界面无效块告警消除