[
https://issues.apache.org/jira/browse/HDFS-12935?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Jianfei Jiang updated HDFS-12935:
---------------------------------
Description:
In HA mode, if one namenode is down, most of functions can still work. When
considering the following two occasions:
(1)nn1 up and nn2 down
(2)nn1 down and nn2 up
These two occasions should be equivalent. However, some of the DFSAdmin
commands will have ambiguous results. The commands can be send successfully to
the up namenode and are always functionally useful only when nn1 is up
regardless of exception (IOException when connecting to the down namenode). See
the following command "hdfs dfsadmin setBalancerBandwidth" which aim to set
balancer bandwidth value for datanodes as an example. It works and all the
datanodes can get the setting values only when nn1 is up. If only nn2 is up,
the command throws exception directly and no datanode get the bandwidth
setting. Approximately ten DFSAdmin commands use the similar logical process
and may be ambiguous.
[root@jiangjianfei01 ~]# hdfs haadmin -getServiceState nn1
active
[root@jiangjianfei01 ~]# hdfs dfsadmin -setBalancerBandwidth 12345
*Balancer bandwidth is set to 12345 for jiangjianfei01/172.17.0.14:9820*
setBalancerBandwidth: Call From jiangjianfei01/172.17.0.14 to
jiangjianfei02:9820 failed on connection exception: java.net.ConnectException:
Connection refused; For more details see:
http://wiki.apache.org/hadoop/ConnectionRefused
[root@jiangjianfei01 ~]# hdfs haadmin -getServiceState nn2
active
[root@jiangjianfei01 ~]# hdfs dfsadmin -setBalancerBandwidth 1234
setBalancerBandwidth: Call From jiangjianfei01/172.17.0.14 to
jiangjianfei01:9820 failed on connection exception: java.net.ConnectException:
Connection refused; For more details see:
http://wiki.apache.org/hadoop/ConnectionRefused
[root@jiangjianfei01 ~]#
was:
In HA mode, if one namenode is down, most of functions can still work. When
considering the following two occasions: (1)nn1 up and nn2 down, (2)nn1 down
and nn2 up. These two occasions should be equivalent. However, some of the
DFSAdmin commands will have ambiguous results. The commands can be send
successfully to the up namenode and are always functionally useful only when
nn1 is up regardless of exception (IOException when connecting to the down
namenode). See the following command "hdfs dfsadmin setBalancerBandwidth" which
aim to set balancer bandwidth value for datanodes as an example. It works and
all the datanodes can get the setting values only when nn1 is up. If only nn2
is up, the command throws exception directly and no datanode get the bandwidth
setting. Approximately ten DFSAdmin commands use the similar logical process
and may be ambiguous.
[root@jiangjianfei01 ~]# hdfs haadmin -getServiceState nn1
active
[root@jiangjianfei01 ~]# hdfs dfsadmin -setBalancerBandwidth 12345
*Balancer bandwidth is set to 12345 for jiangjianfei01/172.17.0.14:9820*
setBalancerBandwidth: Call From jiangjianfei01/172.17.0.14 to
jiangjianfei02:9820 failed on connection exception: java.net.ConnectException:
Connection refused; For more details see:
http://wiki.apache.org/hadoop/ConnectionRefused
[root@jiangjianfei01 ~]# hdfs haadmin -getServiceState nn2
active
[root@jiangjianfei01 ~]# hdfs dfsadmin -setBalancerBandwidth 1234
setBalancerBandwidth: Call From jiangjianfei01/172.17.0.14 to
jiangjianfei01:9820 failed on connection exception: java.net.ConnectException:
Connection refused; For more details see:
http://wiki.apache.org/hadoop/ConnectionRefused
[root@jiangjianfei01 ~]#
> Get ambiguous result for DFSAdmin command in HA mode when only one namenode
> is up
> ---------------------------------------------------------------------------------
>
> Key: HDFS-12935
> URL: https://issues.apache.org/jira/browse/HDFS-12935
> Project: Hadoop HDFS
> Issue Type: Bug
> Components: tools
> Affects Versions: 3.0.0-beta1
> Reporter: Jianfei Jiang
> Fix For: 3.0.0-beta1
>
> Attachments: HDFS_12935.001.patch
>
>
> In HA mode, if one namenode is down, most of functions can still work. When
> considering the following two occasions:
> (1)nn1 up and nn2 down
> (2)nn1 down and nn2 up
> These two occasions should be equivalent. However, some of the DFSAdmin
> commands will have ambiguous results. The commands can be send successfully
> to the up namenode and are always functionally useful only when nn1 is up
> regardless of exception (IOException when connecting to the down namenode).
> See the following command "hdfs dfsadmin setBalancerBandwidth" which aim to
> set balancer bandwidth value for datanodes as an example. It works and all
> the datanodes can get the setting values only when nn1 is up. If only nn2 is
> up, the command throws exception directly and no datanode get the bandwidth
> setting. Approximately ten DFSAdmin commands use the similar logical process
> and may be ambiguous.
> [root@jiangjianfei01 ~]# hdfs haadmin -getServiceState nn1
> active
> [root@jiangjianfei01 ~]# hdfs dfsadmin -setBalancerBandwidth 12345
> *Balancer bandwidth is set to 12345 for jiangjianfei01/172.17.0.14:9820*
> setBalancerBandwidth: Call From jiangjianfei01/172.17.0.14 to
> jiangjianfei02:9820 failed on connection exception:
> java.net.ConnectException: Connection refused; For more details see:
> http://wiki.apache.org/hadoop/ConnectionRefused
> [root@jiangjianfei01 ~]# hdfs haadmin -getServiceState nn2
> active
> [root@jiangjianfei01 ~]# hdfs dfsadmin -setBalancerBandwidth 1234
> setBalancerBandwidth: Call From jiangjianfei01/172.17.0.14 to
> jiangjianfei01:9820 failed on connection exception:
> java.net.ConnectException: Connection refused; For more details see:
> http://wiki.apache.org/hadoop/ConnectionRefused
> [root@jiangjianfei01 ~]#
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
