[
https://issues.apache.org/jira/browse/NIFI-6197?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Lars Francke updated NIFI-6197:
-------------------------------
Resolution: Won't Fix
Status: Resolved (was: Patch Available)
It was decided not to change the current behavior.
The reasoning is that a) an increased default timeout can lead to bad user
experience when the initial connection is fails and takes a long time and b)
people might want a retry loop anyway.
I've therefore closed the PR & will close this issue but I'm happy to revive
the patch should opinions change.
> HBase Client Service uses a bad default for retries
> ---------------------------------------------------
>
> Key: NIFI-6197
> URL: https://issues.apache.org/jira/browse/NIFI-6197
> Project: Apache NiFi
> Issue Type: Bug
> Reporter: Lars Francke
> Assignee: Lars Francke
> Priority: Major
> Time Spent: 4h 50m
> Remaining Estimate: 0h
>
> - This fixes deprecations in HBase 2.0 which will be removed in 3.0. Amongst
> other things the typeByte, sequenceId and tags are now considered internal
> and are not exposed anymore.
> - I also made it so that only either an hbase-site.xml or a ZooKeeper quorum
> need to be provided, previously ZooKeeper Client Port, Client retry and ZK
> Parent Node were mandatory as well but they are not really needed to connect
> to HBase
> - Removed the default for HBase client retries. This means it can still be
> overridden if needed but if not the defaults from either hbase-site.xml or
> HBase itself will be used. HBase Client Service uses a default retry value of
> "1" for {{hbase.client.retries.number}} which is bad. The property is used
> for _all_ *retryable* operations. By setting this to one this makes any
> _normal_ HBase exception (e.g. RegionMoved) fatal.
> - Some minor warnings (e.g. too broad scope, unchecked cast etc.)
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
