[
https://issues.apache.org/jira/browse/HDDS-7187?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17598247#comment-17598247
]
Kaijie Chen commented on HDDS-7187:
-----------------------------------
{quote}
This should be good improvement. The retry logic can be implemented to
XceiverClient layer before giving it up and propagating the exception. What do
you think?
{quote}
The problem is, RATIS/THREE read and maybe other requests are also using
GrpcXceiverClient. I'm not sure if retry is desired at all requests. It's
probably better to switch to another datanode early in the RATIS/THREE read
case.
Should we add a new method like sendCommandAsyncWithRetry?
> EC: Retry failed writes before rewrite to a new block group
> -----------------------------------------------------------
>
> Key: HDDS-7187
> URL: https://issues.apache.org/jira/browse/HDDS-7187
> Project: Apache Ozone
> Issue Type: Sub-task
> Reporter: Kaijie Chen
> Assignee: Kaijie Chen
> Priority: Major
>
> When WriteChunk or PutBlock fails when writing stripe, client should retry
> the operation before rewrite this stripe to a new block. So that the block
> files on disk will be as large as possible.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
