[
https://issues.apache.org/jira/browse/ZOOKEEPER-2449?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Vishal Khandelwal updated ZOOKEEPER-2449:
-----------------------------------------
Description:
As per the current logic if a host has "UnknownHostException" during zookeeper
client object connection and it is never refreshed after that. Incase host
comes back, this class won't try to connect it back.
Ideally in StaticHostProvider.next we end of the list is reached then all of
the host in connection string should be tested and refreshed again. That way
incase host comes back and object stays for a longer duration, realtime status
can be updated and client can take advantage of that
This also benefit the scenario defined
https://issues.apache.org/jira/browse/ZOOKEEPER-2447
was:
As per the current logic if a host has "UnknownHostException" during zookeeper
client object connection and it is never refreshed after that. Incase host
comes back, this class won't try to connect it back.
Ideally in StaticHostProvider.next we end of the list is reached then all of
the host in connection string should be tested and refreshed again. That way
incase host comes back and object stays for a longer duration, realtime status
can be updated and client can take advantage of that
> Refresh the invalid host which got unknownhostexception during connection in
> StaticHostProvider
> -----------------------------------------------------------------------------------------------
>
> Key: ZOOKEEPER-2449
> URL: https://issues.apache.org/jira/browse/ZOOKEEPER-2449
> Project: ZooKeeper
> Issue Type: Improvement
> Reporter: Vishal Khandelwal
> Priority: Minor
>
> As per the current logic if a host has "UnknownHostException" during
> zookeeper client object connection and it is never refreshed after that.
> Incase host comes back, this class won't try to connect it back.
> Ideally in StaticHostProvider.next we end of the list is reached then all of
> the host in connection string should be tested and refreshed again. That way
> incase host comes back and object stays for a longer duration, realtime
> status can be updated and client can take advantage of that
> This also benefit the scenario defined
> https://issues.apache.org/jira/browse/ZOOKEEPER-2447
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
