[
https://issues.apache.org/jira/browse/HADOOP-12067?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14579389#comment-14579389
]
Chris Nauroth commented on HADOOP-12067:
----------------------------------------
Hi Duo. Here are a few comments on this patch.
I'd also appreciate getting a second code review on this one. [~ivanmi] or
[~chuanliu], would you like to take a look?
# When catching {{InterruptedException}}, we need to re-raise the interrupted
signal in case any code that called into this layer relies on seeing the
interrupted status, such as daemon shutdown logic. To do this, call
{{Thread.currentThread().interrupt()}} inside the {{catch}} block. Also,
please use the {{LOG}} if you want to print information about the exception.
{{e.printStackTrace()}} would go to stderr.
# I think the {{continue}} on line 2485 is unnecessary. Can you please remove
it?
# Inside the retry loop, there is some code duplication for {{else}} paths to
{{throw ex}}. I think we can cut one of these code paths by changing the
conditional to this:
{code}
if (ex.getErrorCode().equals(
StorageErrorCode.SERVER_BUSY.toString()) &&
retryCount < copyBlobMaxRetries) {
{code}
# It would be great too if we could eliminate the code duplication caused by
multiple calls to {{dstBlob.startCopyFromBlob}}: once to get started and a
second time in the exception handler. This isn't directly related to the
current patch, but let me know if you have any thoughts on cleaning this up.
> Add exponential retry when copyblob is throttled by Azure storage
> -----------------------------------------------------------------
>
> Key: HADOOP-12067
> URL: https://issues.apache.org/jira/browse/HADOOP-12067
> Project: Hadoop Common
> Issue Type: Bug
> Components: tools
> Affects Versions: 2.7.0
> Reporter: Duo Xu
> Assignee: Duo Xu
> Attachments: HADOOP-12067.01.patch
>
>
> HADOOP-11693 passes exponential retry policy to Azure storage SDK when Azure
> storage throttling happens. However, when I looked at the source code of
> Azure storage SDK. Storage exceptions such as throttling exception are
> non-retryable.
> I would like to add retry in WASB driver instead of depending on Azure
> storage SDK to make sure retry happens when Azure storage throttling fires.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
