每次对logger.info/debug/warn（）调用进行多次调用时，仅使用log4j记录一次

Question

I have a scenario where a particular log message might get printed a lot of times (may be in millions). For example, if we log (using logger.warn() method)for every record with the missing field(s), we might end up logging a lot-cases where input file has a lot of records with missing fields(for example, large files on HDFS). This quickly fills up the disk space.

To avoid this situation, I am trying to log once for every (for example) 1000 records with missing fields. I can implement all of this logic outside of the log4j package, but I was wondering if there is a cleaner way to do this. Ideally, all of this logic would go into the log4j code.

This seems like a commonly encountered problem, but there is hardly any info on this. Any thoughts?

Answer 1

Log4J cant do that out of the box. However you may try to write your own listener. If you want to switch to Logback as your logging framework, there is a Filter called DuplicateMessageFilter that drops messages after a certain repetition. You should really consider this, because that much logging will surely impact on your performance. Logback is configured the same way as Log4J and supports SLF4J out of the box.

Answer 2

You can use a counter and set the log level programmatically. Not the best software design, but enough if you want to do this kind of logging only at one point.

import org.apache.log4j.Level;
import org.apache.log4j.Logger;

public class LogExample {

    private static final Logger LOG = Logger.getLogger(LogExample.class);

    private static final Level DEFAULT_LOG_LEVEL = Level.ERROR;

    public static void main(final String[] args) {
        int count = 0;
        LOG.setLevel(DEFAULT_LOG_LEVEL);
        for (int i = 1; i < 1000000; i++) {
            count++;
            final boolean logInfo = (count % 1000) == 0;
            if (logInfo) {
                LOG.setLevel(Level.INFO);
            }
            LOG.info("test: " + i);
            if (logInfo) {
                LOG.setLevel(DEFAULT_LOG_LEVEL);
            }
        }
    }
}