[
https://issues.apache.org/jira/browse/HADOOP-11387?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14247815#comment-14247815
]
Haohui Mai commented on HADOOP-11387:
-------------------------------------
bq. Have a look at TEZ-1526 to see my earlier concerns with the performance and
memory concerns over using this guava library in Hadoop.
As [~gtCarrera9] pointed out, the cache is only for caching the resolution of
NN by the client-side {{FileSystem}} objects. In common cases there should be
only one or two instances in the cache, thus in my opinion guava cache is good
enough for this use case.
bq. That is not documented in this jira. If there is truly a simplification
jira then this shouldn't block a clean jenkins build. Can you please comment on
this.
I'll update the description accordingly.
> Simplify NetUtils#canonicalizeHost()
> ------------------------------------
>
> Key: HADOOP-11387
> URL: https://issues.apache.org/jira/browse/HADOOP-11387
> Project: Hadoop Common
> Issue Type: Bug
> Reporter: Haohui Mai
> Assignee: Li Lu
> Attachments: HADOOP-11387-121514-1.patch,
> HADOOP-11387-121514-2.patch, HADOOP-11387-121514.patch
>
>
> Currently {{NetUtils#canonicalizeHost}} uses a {{ConcurrentHashMap}} to cache
> the canonicalized hostname.
> {code}
> private static String canonicalizeHost(String host) {
> // check if the host has already been canonicalized
> String fqHost = canonicalizedHostCache.get(host);
> if (fqHost == null) {
> try {
> fqHost = SecurityUtil.getByName(host).getHostName();
> // slight race condition, but won't hurt
> canonicalizedHostCache.put(host, fqHost);
> } catch (UnknownHostException e) {
> fqHost = host;
> }
> }
> return fqHost;
> }
> {code}
> The code can be simplified using {{CacheMap}}.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
