Description
When Hbase client ZooKeeperWatcher gets ConnectionLossException (/hbase/rs or /hbase), it is very hard debug the ZK server log if ZK server has started using log4j INFO level.
When maxClientCnxn limit is reached for a single client (at the socket level), it will be nice to log it in INFO level instead of WARN. It will help hbase clients (Region server, HMaster, and HBase cient lib) to debug the issue in production.
3.4 - src/java/main/org/apache/zookeeper/server/NIOServerCnxnFactory.java 3.3.4 - src/java/main/org/apache/zookeeper/server/NIOServerCnxn.java public void run() { while (!ss.socket().isClosed()) { try { ... ... if (maxClientCnxns > 0 && cnxncount >= maxClientCnxns){ LOG.info("Too many connections from " + ia + " - max is " + maxClientCnxns ); sc.close(); } ... }