[
https://issues.apache.org/jira/browse/HBASE-6387?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13415385#comment-13415385
]
stack commented on HBASE-6387:
------------------------------
@Mikhail That comment might be a bit stale The class javadoc above it says:
{code}
.....Resolves on construction AND on
* deserialization -- since we're internally creating an InetSocketAddress --
* so could end up with different results if the two ends of serialization have
* different resolvers. Be careful where you use it. Should only be used when
* you need to pass an InetSocketAddress across an RPC. Even then its a bad
* idea because of the above resolve issue.
* @deprecated Use {@link InetSocketAddress} or {@link ServerName} or
* a hostname String and port.
{code}
I'd say, yeah, do 89fb only... In trunk we're trying to leave HSA behind.
Good on you.
> Cache DNS lookups in HServerAddress
> -----------------------------------
>
> Key: HBASE-6387
> URL: https://issues.apache.org/jira/browse/HBASE-6387
> Project: HBase
> Issue Type: Improvement
> Reporter: Mikhail Bautin
>
> We have noticed that we rely on DNS lookups in some critical paths by using
> HServerAddress, and Java only seems to be caching DNS data for 30 seconds by
> default. Also, if DNS is down, Java's negative cache of DNS will ensure that
> many successive attempts fail. However, we cannot just increase
> networkaddress.cache.ttl to a large value, because e.g. namenode failover may
> require resolving the same DNS name differently. Therefore I propose that we
> add a DNS lookup cache in HServerAddress.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators:
https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira
