Yes, thanks for the information. I worked around the problem. Thanks, Atul
From: Remko Popma <remko.po...@gmail.com> Date: Thursday, 30 December 2021 at 4:13 PM To: Log4J Users List <log4j-user@logging.apache.org> Subject: Re: RFC5424Layout - How to format timestamp with microsecond precision The class I was talking about is in the core jar. This is a multi release jar (https://docs.oracle.com/javase/10/docs/specs/jar/jar.html#multi-release-jar-files), and the class I was talking about is in the core jar in the META-INF/versions/9 directory. But you have resolved the problem? > On Dec 30, 2021, at 16:55, Atul Pendse <apen...@tintri.com.invalid> wrote: > > BTW.. I am using java 11, with unnamed module. > > Thanks, > Atul > > From: Atul Pendse <apen...@tintri.com.INVALID> > Date: Thursday, 30 December 2021 at 1:24 PM > To: Log4J Users List <log4j-user@logging.apache.org> > Subject: Re: RFC5424Layout - How to format timestamp with microsecond > precision > Sorry, I hit send button prematurely. > > I worked around the issue by explicitly setting Configuration.nanoClock to > SystemNanoClock, only if nano precision is required. Similar to how > PatternParser does it. > > I see below code in PatternParser > > if (converter instanceof NanoTimePatternConverter) { > // LOG4J2-1074 Switch to actual clock if nanosecond timestamps > are required in config. > // LOG4J2-1248 set config nanoclock > if (config != null) { > config.setNanoClock(new SystemNanoClock()); > } > } > > On similar line, I handled my customized Rfc5424Layout > > if ( TIME_PRECISION_MICRO.equalsIgnoreCase( timestampPrecision ) && !( > configuration.getNanoClock() instanceof SystemNanoClock ) ) { > configuration.setNanoClock( new SystemNanoClock() ); > } > > Time precision is a new attribute that I added in my copy of source. > > Thanks, > Atul > > From: Atul Pendse <apen...@tintri.com> > Date: Thursday, 30 December 2021 at 1:16 PM > To: Log4J Users List <log4j-user@logging.apache.org> > Subject: Re: RFC5424Layout - How to format timestamp with microsecond > precision > Ok, I am confused about log4j-core-java9 > I am using below jars > > Log4j-core-2.17.0.jar > Log4j-api-2.17.0.jar > > I don’t see anything like log4j-core-java9-2.17.0.jar on maven. Where can I > find it? > > In my environment, ClockFactory is creating instance of SystemClock, which > looks like below > > package org.apache.logging.log4j.core.util; > /** > * Implementation of the {@code Clock} interface that returns the system time. > */ > public final class SystemClock implements Clock { > > /** > * Returns the system time. > * @return the result of calling {@code System.currentTimeMillis()} > */ > @Override > public long currentTimeMillis() { > return System.currentTimeMillis(); > } > > } > > I see no way it can support nanoseconds. > > From: Remko Popma <remko.po...@gmail.com> > Date: Thursday, 30 December 2021 at 12:53 PM > To: Log4J Users List <log4j-user@logging.apache.org> > Subject: Re: RFC5424Layout - How to format timestamp with microsecond > precision > Atul, > > The nanoClock is not used for the formatted timestamps. It just outputs the > result of System.nanoTime(). It is used for %N or %nano in PatternLayout. > > If you are running on Java 9 or later, then the default clock > is > log4j-core-java9/src/main/java/org/apache/logging/log4j/core/util/SystemClock.java > (Note that there is another class, also called SystemClock, in the same > package, in log4j-core/src.) > > The SystemClock that is loaded on Java 9 and later implements both the > Clock and the PreciseClock interface. > Log4jLogEvent has a field 'instant' of type MutableInstant; this field is > initialized from the Clock. > (There is a Log4jLogEvent constructor where the instant is not initialized > from the Clock, but I believe this is only used in special cases.) > > So, I don't believe that Log4jLogEvent should prevent you from seeing > sub-millisecond timestamps. > Something else must be the problem. > > Can you verify that in your environment, the Log4jLogEvent's CLOCK (the > result of calling ClockFactory.getClock()) implements > org.apache.logging.log4j.core.time.PreciseClock? > > > > > > >> On Thu, Dec 30, 2021 at 4:04 PM Ralph Goers <ralph.go...@dslextreme.com> >> wrote: >> >> I think you want to set >> log4j2.clock=org.apache.logging.log4j.core.time.internal.FixedPreciseClock. >> >> If you don’t like that implementation you can always provide your own >> implementation of PreciseClock and specify it on the system property. >> >> I’m not sure what performance impact that Clock implementation will have. >> >> Ralph >> >> >>> On Dec 29, 2021, at 11:56 AM, Atul Pendse <apen...@tintri.com.INVALID> >> wrote: >>> >>> Thanks for the quick response. It took me some time to get back to this >> due to other priorities. >>> I logged a JIRA as you suggested. >>> >>> Also, meanwhile, I am trying to get it working for our project by >> copying sources and modifying them. >>> I have add a new FixedDateFormat that has microsecond precision with >> RFC5424 date format. I am using the new format in Rfc5424Layout, but I came >> across another problem. >>> >>> It works fine when I run it in a standalone java app, but does not work >> in a web app (like war deployed to tomcat). >>> >>> Standalone java app uses ReusableLogEventFactory to create LogEvent. It >> creates a MutableLogEvent, which generates actual system nano time. >>> >>> In case of web app, log4j uses DefaultLogEventFactory, which creates >> instances of Log4jLogEvent. Log4jLogEvent uses a DummyNanoClock, which >> always defaults nano seconds to value 0. This means, events generated in a >> web app are never going to have micro or nanosecond information. >>> >>> I can override the LogEventFactory by passing system property >> ‘-DLog4jLogEventFactory=org.apache.logging.log4j.core.impl.ReusableLogEventFactory’, >> but I am not really sure if there are any side effects of doing that. Does >> anyone know if it’s okay to use ReusableLogEventFactory for a web app? >>> >>> Thanks, >>> Atul >>> >>> From: Volkan Yazıcı <vol...@yazi.ci> >>> Date: Thursday, 23 December 2021 at 2:29 PM >>> To: Log4J Users List <log4j-user@logging.apache.org> >>> Subject: Re: RFC5424Layout - How to format timestamp with microsecond >> precision >>> Hello Atul, >>> >>> Your investigation of RFC5424 Layout seems to be accurate; the class >>> contains a hardcoded date-time formatting working on >> millisecond-precision >>> epoch-offset timestamps. I guess two things need to happen: >>> >>> 1. A new layout configuration argument where one can override the >>> date-time format pattern. >>> 2. Switching from the hardcoded date-time formatter to Log4j-provided >>> ones; FastDateFormat, FixedDateFormat, etc. I would suggest checking >>> JsonTemplateLayout's InstantFormatter – the most up-to-date one, AFAIC. >>> >>> These better be placed into a JIRA ticket followed by a GitHub PR. >>> >>> In the meantime, you can copy and adapt the source of Rfc5424Layout in >> your >>> project, change the plugin name, and use it there. This can serve you >> until >>> the feature gets implemented and shipped with a release. >>> >>> Kind regards. >>> >>> On Thu, Dec 23, 2021 at 9:38 AM Atul Pendse <apen...@tintri.com.invalid> >>> wrote: >>> >>>> Hi, >>>> >>>> Our application sends syslog messages to a centralized facility which >>>> requires timestamps to be in microsecond format. >>>> Our application was so far using log4j 1.12.5’s Syslog appender to send >>>> syslog messages, which used to format timestamps with microsecond >> precision. >>>> >>>> We are now upgrading to log4j 2.17.0, and don’t see any way to get >>>> microseconds included in timestamp for syslog. >>>> >>>> I am trying to log messages using SyslogAppender with RFC5424 layout. >>>> I see that RFC5424Layout.java restricts timestamp to millisecond >> precision >>>> (e.g. 2021-12-22T22:54:33.889-08:00). >>>> >>>> RFC5424 specification also mentions support for microsecond precision. >>>> Here is some text from >>>> https://datatracker.ietf.org/doc/html/rfc5424#section-6.2.3 >>>> >>>> Example 4 >>>> >>>> 2003-08-24T05:14:15.000003-07:00 >>>> >>>> This represents 24 August 2003 at 05:14:15am, 3 microseconds into the >>>> next second. The microsecond resolution is indicated by the >>>> additional digits in TIME-SECFRAC. The timestamp indicates that its >>>> local time is -7 hours from UTC. This timestamp might be created in >>>> the US Pacific time zone during daylight savings time. >>>> >>>> >>>> Is there a way to format timestamp with microsecond precision with >>>> RFC5424Layout? Or is there an alternative way to get timestamp formatted >>>> with microsecond precision for SyslogAppender? >>>> >>>> Any help would be highly appreciated. >>>> >>>> Thanks, >>>> Atul Pendse >>>> >> >> >> --------------------------------------------------------------------- >> To unsubscribe, e-mail: log4j-user-unsubscr...@logging.apache.org >> For additional commands, e-mail: log4j-user-h...@logging.apache.org >> >>
