[
https://issues.apache.org/jira/browse/NUTCH-2896?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17579739#comment-17579739
]
ASF GitHub Bot commented on NUTCH-2896:
---------------------------------------
sebastian-nagel commented on PR #697:
URL: https://github.com/apache/nutch/pull/697#issuecomment-1215088177
@lewismc: I've linked this issue on a new wiki page
[ProtocolImplementations](https://cwiki.apache.org/confluence/display/NUTCH/ProtocolImplementations).
Merging now - the connection pool is tested since almost one year now.
> Protocol-okhttp: make connection pool configurable
> --------------------------------------------------
>
> Key: NUTCH-2896
> URL: https://issues.apache.org/jira/browse/NUTCH-2896
> Project: Nutch
> Issue Type: Improvement
> Components: plugin, protocol
> Affects Versions: 1.18
> Reporter: Sebastian Nagel
> Priority: Major
> Fix For: 1.19
>
>
> [OkHttp's
> ConnectionPool|https://square.github.io/okhttp/3.x/okhttp/okhttp3/ConnectionPool.html]
> "holds up to 5 idle connections which will be evicted after 5 minutes of
> inactivity." A pool of this size is suitable for site crawls but not for
> larger crawls over many different sites / hosts.
> Note: in the current version (4.9.1) the connection pool is implemented as a
> [linked
> queue|https://github.com/square/okhttp/blob/parent-4.9.1/okhttp/src/main/kotlin/okhttp3/internal/connection/RealConnectionPool.kt#L83].
> In order to scale beyond pool sizes exceeding 1000 we need to use a set of
> clients each with its own connection pool.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
