You are viewing a plain text version of this content. The canonical link for it is here.
Posted to common-dev@hadoop.apache.org by Scott Forman <ls...@gmail.com> on 2013/05/10 17:53:28 UTC

Web UI for Active and Standby NN do not agree on safe mode setting

When safe mode is entered using the dfsadmin CLI command, only one of the
Web UI's for the Namenodes in an HA environment seems to show that safe
mode is on.  But retrieving the safe mode setting through the CLI (i.e.
running "hdfs dfsadmin -safemode get") seems to display the correct setting
when run from any node in the cluster.

I noticed this problem when upgrading from a non-HA environment to an HA
environment, and in this case, the NN in the pre-upgraded system always
shows the correct setting, regardless if it is active or standby, while the
newly added NN never displays that safe mode is on.

The dfsadmin command uses DistributedFileSystem to get the setting, while
the Web UI seems to use FSNamesystem.

Is this a bug, or do both UI's need to be consulted to determine the safe
mode setting?

Thanks,
Scott

Re: Web UI for Active and Standby NN do not agree on safe mode setting

Posted by "Aaron T. Myers" <at...@cloudera.com>.
Hi Scott,

This is actually working as intended. NN safemode is not a system-wide
state - it can be set independently on both the active and standby NNs.
DistributedFileSystem is indeed "doing the right thing" in that it will
return the safemode value of the active NN, but on the web UI each NN
should display its own safemode state, which might differ from the other's.
You might take a look at the discussion/patch in
https://issues.apache.org/jira/browse/HDFS-3507 for some more background
here.

Best,
Aaron


--
Aaron T. Myers
Software Engineer, Cloudera


On Fri, May 10, 2013 at 8:53 AM, Scott Forman <ls...@gmail.com>wrote:

> When safe mode is entered using the dfsadmin CLI command, only one of the
> Web UI's for the Namenodes in an HA environment seems to show that safe
> mode is on.  But retrieving the safe mode setting through the CLI (i.e.
> running "hdfs dfsadmin -safemode get") seems to display the correct setting
> when run from any node in the cluster.
>
> I noticed this problem when upgrading from a non-HA environment to an HA
> environment, and in this case, the NN in the pre-upgraded system always
> shows the correct setting, regardless if it is active or standby, while the
> newly added NN never displays that safe mode is on.
>
> The dfsadmin command uses DistributedFileSystem to get the setting, while
> the Web UI seems to use FSNamesystem.
>
> Is this a bug, or do both UI's need to be consulted to determine the safe
> mode setting?
>
> Thanks,
> Scott
>