[
https://issues.apache.org/jira/browse/HDFS-8809?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Jing Zhao updated HDFS-8809:
----------------------------
Attachment: HDFS-8809.000.patch
There is another issue here, which I think exists before HDFS-8215. When
"-OPENFORWRITE" is enabled, an UC block is still treated as missing/corrupted,
since {{countNodes}} only checks the triplets inside of the BlockInfo and thus
the liveReplicas is usually 0 for a UC block.
The fix can be to ignore the check if the block is the last one and it's UC.
Theoretically the penultimate block can also be in the committed state and with
0 reported replica yet, but maybe we do not need to handle this part here.
> HDFS fsck reports HBase WALs files (under construction) as "CORRUPT" (missing
> blocks) when HBase is running
> -----------------------------------------------------------------------------------------------------------
>
> Key: HDFS-8809
> URL: https://issues.apache.org/jira/browse/HDFS-8809
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: tools
> Affects Versions: 2.7.0
> Environment: Hadoop 2.7.1 and HBase 1.1.1, on SUSE11sp3 (other
> Linuxes not tested, probably not platform-dependent). This did NOT happen
> with Hadoop 2.4 and HBase 0.98.
> Reporter: Sudhir Prakash
> Assignee: Jing Zhao
> Attachments: HDFS-8809.000.patch
>
>
> Whenever HBase is running, the "hdfs fsck /" reports four hbase-related
> files in the path "hbase/data/WALs/" as CORRUPT. Even after letting the
> cluster sit idle for a couple hours, it is still in the corrupt state. If
> HBase is shut down, the problem goes away. If HBase is then restarted, the
> problem recurs. This was observed with Hadoop 2.7.1 and HBase 1.1.1, and did
> NOT happen with Hadoop 2.4 and HBase 0.98.
> {code}
> hades1:/var/opt/teradata/packages # su hdfs
> hdfs@hades1:/var/opt/teradata/packages> hdfs fsck /
> Connecting to namenode via
> http://hades1.labs.teradata.com:50070/fsck?ugi=hdfs&path=%2F
> FSCK started by hdfs (auth:SIMPLE) from /39.0.8.2 for path / at Wed Jun 24
> 20:40:17 GMT 2015
> ...
> /apps/hbase/data/WALs/hades4.labs.teradata.com,16020,1435168292684/hades4.labs.teradata.com%2C16020%2C1435168292684.default.1435175500556:
> MISSING 1 blocks of total size 83 B.
> /apps/hbase/data/WALs/hades5.labs.teradata.com,16020,1435168290466/hades5.labs.teradata.com%2C16020%2C1435168290466..meta.1435175562144.meta:
> MISSING 1 blocks of total size 83 B.
> /apps/hbase/data/WALs/hades5.labs.teradata.com,16020,1435168290466/hades5.labs.teradata.com%2C16020%2C1435168290466.default.1435175498500:
> MISSING 1 blocks of total size 83 B.
> /apps/hbase/data/WALs/hades6.labs.teradata.com,16020,1435168292373/hades6.labs.teradata.com%2C16020%2C1435168292373.default.1435175500301:
> MISSING 1 blocks of total size 83
> B..................................................................................................
> ....................................................................................................
> ....................................................................................................
> ........................................................................................Status:
> CORRUPT
> Total size: 723977553 B (Total open files size: 332 B)
> Total dirs: 79
> Total files: 388
> Total symlinks: 0 (Files currently being written: 5)
> Total blocks (validated): 387 (avg. block size 1870743 B) (Total open
> file blocks (not validated): 4)
> ********************************
> UNDER MIN REPL'D BLOCKS: 4 (1.0335917 %)
> dfs.namenode.replication.min: 1
> CORRUPT FILES: 4
> MISSING BLOCKS: 4
> MISSING SIZE: 332 B
> ********************************
> Minimally replicated blocks: 387 (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: 3
> Number of racks: 1
> FSCK ended at Wed Jun 24 20:40:17 GMT 2015 in 7 milliseconds
> The filesystem under path '/' is CORRUPT
> hdfs@hades1:/var/opt/teradata/packages>
> {code}
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
