[
https://issues.apache.org/jira/browse/OOZIE-2457?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15948214#comment-15948214
]
Robert Kanter commented on OOZIE-2457:
--------------------------------------
Hi [~jfilipiak]. We've discussed doing something like that for OOZIE-1561
which is trying to fix an issue where logs are not HA when Oozie HA is enabled
(because they're on individual Oozie servers). The guy who was working on that
isn't working on Oozie anymore, so it hasn't been active in a long time.
There's some complications with storing the logs in HDFS; please take a look at
what's been discussed there. If you'd like to take this up, I'm sure people
would appreciate it. In any case, let's move this discussion to OOZIE-1561 and
leave the regex cpu issue for here.
> Oozie log parsing regex consume more than 90% cpu
> -------------------------------------------------
>
> Key: OOZIE-2457
> URL: https://issues.apache.org/jira/browse/OOZIE-2457
> Project: Oozie
> Issue Type: Bug
> Reporter: Satish Subhashrao Saley
> Assignee: Satish Subhashrao Saley
> Priority: Blocker
> Fix For: 5.0.0
>
> Attachments: OOZIE-2457-1.patch, OOZIE-2457-2.patch,
> OOZIE-2457-3.patch, OOZIE-2457-4.patch, OOZIE-2457-5.patch, OOZIE-2457-6.patch
>
>
> http-0.0.0.0-4080-26 TID=62215 STATE=RUNNABLE CPU_TIME=1992 (92.59%)
> USER_TIME=1990 (92.46%) Allocted: 269156584
> java.util.regex.Pattern$Curly.match0(Pattern.java:4170)
> java.util.regex.Pattern$Curly.match(Pattern.java:4132)
> java.util.regex.Pattern$GroupHead.match(Pattern.java:4556)
> java.util.regex.Matcher.match(Matcher.java:1221)
> java.util.regex.Matcher.matches(Matcher.java:559)
> org.apache.oozie.util.XLogFilter.matches(XLogFilter.java:136)
>
> org.apache.oozie.util.TimestampedMessageParser.parseNextLine(TimestampedMessageParser.java:145)
>
> org.apache.oozie.util.TimestampedMessageParser.increment(TimestampedMessageParser.java:92)
> Regex
> {code}
> (.* USER\[[^\]]*\] GROUP\[[^\]]*\] TOKEN\[[^\]]*\] APP\[[^\]]*\]
> JOB\[0000000-150625114739728-oozie-puru-W\] ACTION\[[^\]]*\] .*)
> {code}
> For single line parsing we use two regex.
> 1.
> {code}
> public ArrayList<String> splitLogMessage(String logLine) {
> Matcher splitter = SPLITTER_PATTERN.matcher(logLine);
> if (splitter.matches()) {
> ArrayList<String> logParts = new ArrayList<String>();
> logParts.add(splitter.group(1));// timestamp
> logParts.add(splitter.group(2));// log level
> logParts.add(splitter.group(3));// Log Message
> return logParts;
> }
> else {
> return null;
> }
> }
> {code}
> 2.
> {code}
> public boolean matches(ArrayList<String> logParts) {
> if (getStartDate() != null) {
> if (logParts.get(0).substring(0,
> 19).compareTo(getFormattedStartDate()) < 0) {
> return false;
> }
> }
> String logLevel = logParts.get(1);
> String logMessage = logParts.get(2);
> if (this.logLevels == null ||
> this.logLevels.containsKey(logLevel.toUpperCase())) {
> Matcher logMatcher = filterPattern.matcher(logMessage);
> return logMatcher.matches();
> }
> else {
> return false;
> }
> }
> {code}
> Also there is repetitive parsing for same log message in
> {code}
> private String parseTimestamp(String line) {
> String timestamp = null;
> ArrayList<String> logParts = filter.splitLogMessage(line);
> if (logParts != null) {
> timestamp = logParts.get(0);
> }
> return timestamp;
> }
> {code}
> where the {{line}} has already parsed using regex and we already know the
> {{logParts}} if any.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
