I'll give a brief description on the setup so that hopefully you have a better
picture.
We want to capture a specific type of event, let's say a business event, from
all the applications on the box and get them to a central location. We're also
looking ahead to a possible PaaS environment and thinking that writing to log
files might not be an option. Our solution is that all the applications on the
box will use our domain socket appender to write these business events into.
In addition, there will be a daemon on this box listening on the domain socket.
It will buffer, compress and send the events to a central location. Even if
we don't want to take into consideration the PaaS environment and thus can
consider writing to local files (as I guess that's the most common logging
solution and has its benefits), there might be issues there. For instance,
let's say we have a process which picks up log files and sends them centrally.
What happens when new applications are installed on the box? We would probably
have to update the configuration of this "log file consolidator" to watch new
folder locations. That might be a bit of a maintenance nightmare.
If the domain socket appender runs into issues, like it can't open the socket,
we need for someone to look into it. I guess the thinking is that if we get
these ERROR/INFO events to a central location we can have some monitoring there
to address these issues.
We also, currently, are failing over the event sent to the domain socket
appender to the domain socket appender's logger, which I guess is what you're
referring to by "Routing events being processed by an Appender through a Logger
is a bad idea.". While that is happening now, that's not the main question I
have here. It's about how to send INFO - ERROR to one location, or possibly
two, and have DEBUG only go to the status logger.
There is push back on "failing over" the events from the domain socket appender
so that code might get pulled out.
Thanks,
Nick
> Subject: Re: status appender?
> From: ralph.go...@dslextreme.com
> Date: Mon, 14 Dec 2015 10:56:11 -0700
> To: log4j-user@logging.apache.org
>
> I don’t think I understand - “we’re writing to our own logger within our
> DomainSocketAppender”. The status logger in your appender should only be
> logging errors or other events that occur within that Appender, which
> normally would be nothing. Routing events being processed by an Appender
> through a Logger is a bad idea.
>
> Ralph
>
>
>
> > On Dec 14, 2015, at 10:17 AM, Nicholas Duane <nic...@msn.com> wrote:
> >
> > Thanks. Yes, that's the one option I was thinking of. Adding a console or
> > file appender to our logger. However, the actual class in question is our
> > DomainSocketAppender. This is somewhat related to a previous question I
> > asked about capturing events from our appenders centrally. I guess in
> > general, the advice is to log to the status logger in your log4j2 classes.
> > We want to capture INFO - ERROR messages centrally so we're writing to our
> > own logger within our DomainSocketAppender. It would be nice to have all
> > other events, DEBUG and less specific (or maybe just all events), go to the
> > status logger and let the application team decide what they want to do with
> > status logger events.
> >
> > I was just thinking that one other solution would be to log all events
> > within our DomainSocketsAppender to both its private logger and the status
> > logger, thus somewhat removing the "routing" within the code. Our filter
> > on the http appender already filters out anything less specific than INFO.
> >
> > Thanks,
> > Nick
> >
> >> Subject: Re: status appender?
> >> From: ralph.go...@dslextreme.com
> >> Date: Mon, 14 Dec 2015 09:32:51 -0700
> >> To: log4j-user@logging.apache.org
> >>
> >> First, what you are wanting to do is, in fact, pretty normal. However, by
> >> default the StatusLogger that Log4j uses for its components doesn’t use an
> >> Appender. Although the API is the same the internals of the StatusLogger
> >> are actually quite different than the “normal” Log4j implementation. That
> >> said, the StatusLogger normally just writes to the console. I actually
> >> doubt that that is where you want your debug events to go. Most people
> >> prefer them to go to a rolling file.
> >>
> >> To accomplish what you want you just need to set up filtering in your
> >> configuration so that the FATAL-INFO events go to one Appender and the
> >> DEBUG and TRACE events go to another Appender.
> >>
> >> Ralph
> >>
> >>> On Dec 14, 2015, at 8:34 AM, Nicholas Duane <nic...@msn.com> wrote:
> >>>
> >>> I'm curious if there is such a thing as a StatusAppender in log4j2 which,
> >>> as you would guess, is the appender the StatusLogger would use?
> >>>
> >>> Here's what I'm trying to solve, I think.
> >>>
> >>> I've been telling other developers I work with that a piece of code
> >>> should only write to a single logger. The reason for this, in my mind,
> >>> is that if a piece of code writes to more than one logger then it
> >>> essentially has routing logic in it and I would rather have the routing
> >>> in the configuration. For example:
> >>>
> >>> try
> >>> {
> >>> logger1.info(...);
> >>> .
> >>> .
> >>> .
> >>> logger2.debug(...);
> >>> }
> >>> catch(Exception e)
> >>> {
> >>> logger1.error(...);
> >>> }
> >>>
> >>> The above code is sending debug events to a different logger than the
> >>> rest of the events it raises. I would rather have the code send all
> >>> events to a single logger and control where those events are routed via
> >>> the configuration. Feel free to let me know whether this is in line with
> >>> logging principles.
> >>>
> >>> So here's the problem. We've got some code which writes events to its
> >>> logger. We want to capture these events centrally so we're sending them
> >>> to a central location via an HTTP appender. We want to do this only for
> >>> FATAL - INFO events, so we're not expecting a huge load. DEBUG events
> >>> however, we'd like to send to the same location as the status logger. We
> >>> can, of course, just add a console appender for DEBUG events but that
> >>> would have to be controlled separately from the status logger and ideally
> >>> it would be nice to just piggy back on the status logger. We could have
> >>> this code write to its private logger and the status logger for DEBUG
> >>> events, but then we get into the routing issue I mentioned above. So I'm
> >>> wondering, is there such a thing as a StatusAppender?
> >>>
> >>> Thanks,
> >>> Nick
> >>>
> >>
> >>
> >>
> >> ---------------------------------------------------------------------
> >> To unsubscribe, e-mail: log4j-user-unsubscr...@logging.apache.org
> >> For additional commands, e-mail: log4j-user-h...@logging.apache.org
> >>
> >
>
>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: log4j-user-unsubscr...@logging.apache.org
> For additional commands, e-mail: log4j-user-h...@logging.apache.org
>
