Critical Description
We run tika in server mode via:
/usr/java/jdk/bin/java -Dlog4j.app.name=-server -Djavax.xml.soap.MessageFactory=com.sun.xml.messaging.saaj.soap.ver1_1.SOAPMessageFactory1_1Impl -Dfile.encoding=UTF-8 -Djava.net.preferIPv4Stack=true -server -Xms256M -Xmx768M -XX:+HeapDumpOnOutOfMemoryError -XX:HeapDumpPath=/var/log/oom/content-extractor-8983.dump.1 -server -Xms500M -Xmx500M -jar /opt/tika/tika-app-1.1.jar --text --encoding=UTF-8 --server 8983
Our client talks to this over port 8983. We pass data via the socket and get the responses back. However, sometimes, tika will get into a bad state and stop responding.
When this happens, we see this in the logs over and over.
2012-05-24_20:12:33.88573 Caused by: java.lang.NullPointerException
2012-05-24_20:12:33.88576 at org.apache.tika.sax.XHTMLContentHandler.lazyEndHead(XHTMLContentHandler.java:157)
2012-05-24_20:12:33.88580 at org.apache.tika.sax.XHTMLContentHandler.startElement(XHTMLContentHandler.java:237)
2012-05-24_20:12:33.88584 at org.apache.tika.sax.XHTMLContentHandler.startElement(XHTMLContentHandler.java:274)
2012-05-24_20:12:33.88589 at org.apache.tika.parser.microsoft.WordExtractor.handleParagraph(WordExtractor.java:186)
2012-05-24_20:12:33.88593 at org.apache.tika.parser.microsoft.WordExtractor.parse(WordExtractor.java:97)
2012-05-24_20:12:33.88597 at org.apache.tika.parser.microsoft.OfficeParser.parse(OfficeParser.java:185)
2012-05-24_20:12:33.88602 at org.apache.tika.parser.microsoft.OfficeParser.parse(OfficeParser.java:160)
2012-05-24_20:12:33.88606 at org.apache.tika.parser.CompositeParser.parse(CompositeParser.java:242)
2012-05-24_20:12:33.88611 ... 4 more
2012-05-24_20:12:49.28441 org.apache.tika.exception.TikaException: Unexpected RuntimeException from org.apache.tika.parser.microsoft.OfficeParse
r@6906daba
2012-05-24_20:12:49.28458 at org.apache.tika.parser.CompositeParser.parse(CompositeParser.java:244)
2012-05-24_20:12:49.28466 at org.apache.tika.parser.CompositeParser.parse(CompositeParser.java:242)
2012-05-24_20:12:49.28477 at org.apache.tika.parser.AutoDetectParser.parse(AutoDetectParser.java:120)
2012-05-24_20:12:49.28489 at org.apache.tika.cli.TikaCLI$OutputType.process(TikaCLI.java:130)
2012-05-24_20:12:49.28497 at org.apache.tika.cli.TikaCLI$TikaServer$1.run(TikaCLI.java:735)
2012-05-24_20:12:49.28509 Caused by: java.lang.NullPointerException
2012-05-24_20:12:49.28516 at org.apache.tika.sax.XHTMLContentHandler.lazyEndHead(XHTMLContentHandler.java:157)
2012-05-24_20:12:49.28524 at org.apache.tika.sax.XHTMLContentHandler.startElement(XHTMLContentHandler.java:237)
2012-05-24_20:12:49.28532 at org.apache.tika.sax.XHTMLContentHandler.startElement(XHTMLContentHandler.java:274)
2012-05-24_20:12:49.28541 at org.apache.tika.parser.microsoft.WordExtractor.handleParagraph(WordExtractor.java:186)
2012-05-24_20:12:49.28550 at org.apache.tika.parser.microsoft.WordExtractor.parse(WordExtractor.java:97)
2012-05-24_20:12:49.28558 at org.apache.tika.parser.microsoft.OfficeParser.parse(OfficeParser.java:185)
2012-05-24_20:12:49.28565 at org.apache.tika.parser.microsoft.OfficeParser.parse(OfficeParser.java:160)
2012-05-24_20:12:49.28577 at org.apache.tika.parser.CompositeParser.parse(CompositeParser.java:242)
2012-05-24_20:12:49.28585 ... 4 more
We have tried to figure out what causes this with no success. We only know that once the server gets into this state, there is no recourse but to restart the tika service.
Other instances of tika we have running in the test environment continue to work. There is some combination of content or work that causes
tika to destabilize. Our working theory is that perhaps tika server is not thread safe and that may be causing this behavior.