[
https://issues.apache.org/jira/browse/HDFS-12410?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16166887#comment-16166887
]
Anu Engineer commented on HDFS-12410:
-------------------------------------
bq. Pushing a config with an unsupported/misspelled type... losing the nodes
with a clear error is easier to debug
I can certainly see both sides of the argument.
1. *Fail fast argument* -- Hadoop should have the smarts to fail when a config
is wrong. As I said I might have mistyped "disc" and now hundreds of disks are
not working, since some "disks" work the cluster appears to be working. In that
case, I would rather that the config failed on me, instead of the mistake being
propagated to thousands of machines. If I am pushing a config with an error, it
is better to fail and easy for me to fix that as soon as I push the deploy
config button.
2. *Postel's Law* - "be conservative in what you do, be liberal in what you
accept from others." I do see that you are arguing that we can be liberal in
what we accept in the config. Unfortunately, when you ignore a storage
configuration option, and you are a file system, It ceases to be conservative
in my opinion.
So I am still inclined to argue that the current behavior is very much what you
want.
> Ignore unknown StorageTypes
> ---------------------------
>
> Key: HDFS-12410
> URL: https://issues.apache.org/jira/browse/HDFS-12410
> Project: Hadoop HDFS
> Issue Type: Task
> Components: datanode, fs
> Reporter: Chris Douglas
> Priority: Minor
>
> A storage configured with an unknown type will cause runtime exceptions.
> Instead, these storages can be ignored/skipped.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
