HalfLegend created HDFS-12694:
---------------------------------
Summary: Wrong data node registered in docker
Key: HDFS-12694
URL: https://issues.apache.org/jira/browse/HDFS-12694
Project: Hadoop HDFS
Issue Type: Bug
Affects Versions: 2.8.1, 2.7.4, 2.8.0
Environment: There hosted machines and each one have a docker.
Hosted machines: host1, host2, host3
dockers : docker1, docker2, docker3
I installed HDFS in each docker. (version 2.8.1)
Docker1 and Docker2 have name node.
Docker1,2,3 all have data node.
I have a weave network of 10.240.1.0/24
Somebody else installed another HDFS in hosted machine outside docker. （Theirs
version is 2.7.0）
Reporter: HalfLegend
Somebody else installed another HDFS in hosted machine outside the dockers.
But the data node in the hosted machine appears in the docker machine.
| Node | Http Address | Capacity | Blocks | Block pool used | Version|
| docker1:50010 (10.240.1.101:50010) | docker1:50075 | 931.06 GB | 8420 |
116.64 GB (12.53%) | 2.8.1 |
| docker2:50010 (10.240.1.102:50010) | docker2:50075 | 916.77 GB | 8420 |
116.64 GB (12.72%) | 2.8.1 |
| docker3:50010 (10.240.1.64:50010) | docker3:50075 | 916.77 GB | 0 | 28 KB
(0%) | 2.8.1 |
The IP address of docker3 should be 10..240.1.103, but it is 10.240.1.64 here
which is the IP address of host3. They also have a data node running on host3.
We can see in docker3, the used blocks is 0, no data is added here. The
replication is 3.
If I use fsck to check the files, many replication blocks are missing. I think
it is because of the IP address.
Further more, if I stop the data node in docker3, the heart beat will stop.
This is correct.
So it is a weird bug.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
---------------------------------------------------------------------
To unsubscribe, e-mail: hdfs-dev-unsubscr...@hadoop.apache.org
For additional commands, e-mail: hdfs-dev-h...@hadoop.apache.org
