[NUTCH-2625] ProtocolFactory.getProtocol(url) may create multiple plugin instances

• Type: Improvement
• Status: Closed
• Priority: Minor
• Resolution: Fixed
• Affects Version/s: 1.15
• Fix Version/s: 1.16
• Component/s: protocol
• Assignee: Sebastian Nagel
• Reporter: Sebastian Nagel
• Created: 24/Jul/18 14:18
• Updated: 28/Jan/21 13:56
• Resolved: 20/Oct/18 17:13

Description

The method ProtocolFactory.getProtocol(URL url) may create unnecessarily multiple instances of protocol plugins given the same configuration. The following snippets from a Fetcher using 100 FetcherThreads show that the setConf(conf) method of the protocol-okhttp plugin is called 100 times (once for each thread):

2018-07-12 12:04:32,811 INFO [main] org.apache.nutch.fetcher.FetcherThread: FetcherThread 1 Using queue mode : byHost
... (skipped 98 repeated messages)
2018-07-12 12:04:33,136 INFO [main] org.apache.nutch.fetcher.FetcherThread: FetcherThread 1 Using queue mode : byHost

...

2018-07-12 12:04:37,493 INFO [FetcherThread] org.apache.nutch.protocol.RobotRulesParser: robots.txt whitelist not configured.
2018-07-12 12:04:37,493 INFO [FetcherThread] org.apache.nutch.protocol.okhttp.OkHttp: http.proxy.host = null
...
2018-07-12 12:04:37,494 INFO [FetcherThread] org.apache.nutch.protocol.okhttp.OkHttp: http.enable.cookie.header = false

... (skipped 98 blocks of repeated messages)

2018-07-12 12:04:39,080 INFO [FetcherThread] org.apache.nutch.protocol.RobotRulesParser: robots.txt whitelist not configured.
2018-07-12 12:04:39,080 INFO [FetcherThread] org.apache.nutch.protocol.okhttp.OkHttp: http.proxy.host = null
...
2018-07-12 12:04:39,080 INFO [FetcherThread] org.apache.nutch.protocol.okhttp.OkHttp: http.enable.cookie.header = false

The method ProtocolFactory.getProtocol(URL url) is synchronized, however each FetcherThread holds its own instance of the ProtocolFactory.

Issue Links

• duplicates: NUTCH-2253 ProtocolFactory still not thread-safe
• links to: GitHub Pull Request #368