[
https://issues.apache.org/jira/browse/HADOOP-11110?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14150262#comment-14150262
]
Andrew Wang commented on HADOOP-11110:
--------------------------------------
Hi Arun, this looks great. I just have a few small comments:
- KeyShell, I notice that we print the success message before flushing in
various places. Should these prints be moved down? I think we wouldn't see this
when testing with the KMS since it always flushes implicitly, but we might when
using JKS.
- FailureInjectingJKSP, could we make the "failjceks" string a public constant
like "jceks" is in JKSP? We can also use JKSP#SCHEME_NAME rather than
hardcoding "jceks" again.
Test:
- Some lines longer than 80 chars
- "faulre furing" is in two comments, typo ;)
- Rather than the wrapper that checks the getClass() is FIJKSP, we could use
KeyProviderFactory#get to get explicitly a failjceks. This is more of a sure
thing, and also we'd definitely not skip the test if somehow what we get out is
not a FIJKSP.
> JavaKeystoreProvider should not report a key as created if it was not flushed
> to the backing file
> -------------------------------------------------------------------------------------------------
>
> Key: HADOOP-11110
> URL: https://issues.apache.org/jira/browse/HADOOP-11110
> Project: Hadoop Common
> Issue Type: Bug
> Affects Versions: 2.5.0
> Reporter: Andrew Wang
> Assignee: Arun Suresh
> Attachments: HADOOP-11110.1.patch
>
>
> Testing with the KMS backed by JKS reveals the following:
> {noformat}
> [root@dlo-4 ~]# hadoop key create testkey -provider
> kms://http@localhost:16000/kms
> testkey has not been created. Mkdirs failed to create file:xxxxx
> ....<stack trace>....
> [root@dlo-4 ~]# hadoop key list -provider kms://http@localhost:16000/kms
> Listing keys for KeyProvider:
> KMSClientProvider[http://localhost:16000/kms/v1/]
> testkey
> {noformat}
> The JKS still has the key in memory and serves it up, but will disappear if
> the KMS is restarted since it's not flushed to the file.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
