hadoop-hdfs-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Anu Engineer (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HDFS-12420) Disable Namenode format when data already exists
Date Wed, 13 Sep 2017 17:47:00 GMT

    [ https://issues.apache.org/jira/browse/HDFS-12420?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16165023#comment-16165023

Anu Engineer commented on HDFS-12420:

bq. Don't we already have the y/n check when data exists? Why do we need another? 
We do, but the fact that it not very clear with lots of other text on the screen was pointed
out by a cluster owner, who was visibly distressed. 

We are just trying to avoid losing data by operator mistake. I thought that you might have
a concern with automation that is why I flagged it for your consideration. Let me try to understand
that a bit more, do you think people automate formatting the clusters? if they do, then preventing
accidental data loss is all the more important.

>From an HDFS user hat on,  I think this is a good improvement to have. I would expect
HDFS to refuse to format a cluster with data. But from a sysadmin/developer hat on, I do like
that fact that I can format a cluster with data. I do that when I test and develop. 

So in my mind, the question boils down to easier dev/ops cycles vs. user safety. The reason
why this is filed for 3.0 is that it might be our last opportunity to make this change.

bq. Completely breaks automation. Automation MUST work. 
I see that you are voting with the devops hat on, and I do not disagree. But this is a place
where breaking the automation might avoid a disaster for some poor user. One more data point,
this JIRA is based on real feedback from a real large cluster.  I am not apologizing for sloppy
operation but trying to understand what we can do to prevent a user from making such a mistake.

I am presuming (please correct me if I am wrong) that you are not objecting to the change
or the intent per se, but more about the fact that we are out right refusing to format a cluster
with Namenode metadata. Do you think adding a flag which says *-DothisIamReallySmart* address
the automation concern?

> Disable Namenode format when data already exists
> ------------------------------------------------
>                 Key: HDFS-12420
>                 URL: https://issues.apache.org/jira/browse/HDFS-12420
>             Project: Hadoop HDFS
>          Issue Type: Improvement
>            Reporter: Ajay Kumar
>            Assignee: Ajay Kumar
>         Attachments: HDFS-12420.01.patch, HDFS-12420.02.patch
> Disable NameNode format to avoid accidental formatting of Namenode in production cluster.
If someone really wants to delete the complete fsImage, they can first delete the metadata
dir and then run {code} hdfs namenode -format{code} manually.

This message was sent by Atlassian JIRA

To unsubscribe, e-mail: hdfs-issues-unsubscribe@hadoop.apache.org
For additional commands, e-mail: hdfs-issues-help@hadoop.apache.org

View raw message