[jira] [Commented] (ATLAS-885) Excessive periodic logs to application.log

["Madhan Neethiraj (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/ATLAS-885?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15321704#comment-15321704 ] 

Madhan Neethiraj commented on ATLAS-885:
----------------------------------------

These logs were generated from Titan's interaction with HBase. HBaseStoreManager.getLocalKeyPartition() instantiates HTable object {{table = new HTable(hconf, tableName);}}. Per [~enis] (HBase committer), this is an expensive operation involving creation of TCP/IP connections (to zookeeper and meta server) and a large number of threads. He suggested to get the table from the connection object, instead of creating HTable directly. Verified that using the suggested approach stopped the excessive logs. I will attach the patch shortly.

Also, the logic used to determine if the storage is local or not, in HBaseStoreManager.getDeployment() looks incorrect. Even if one of the regions of the table exist in the localhost, this method returns the deployment as 'local'. I am not sure how Titan uses the value returned from HBaseStoreManager.getDeployment() and what the impact of wrongly returning as local. In the attache patch, HBaseStoreManager.getDeployment() has been updated to return 'remote' - which should be the case for most production deployments. If this causes any significant performance overhead in single-node deployments, we can revise this logic.

These updates are in Titan implementation and we should have Titan community to review/consider this fix for appropriate versions.

> Excessive periodic logs to application.log
> ------------------------------------------
>
>                 Key: ATLAS-885
>                 URL: https://issues.apache.org/jira/browse/ATLAS-885
>             Project: Atlas
>          Issue Type: Bug
>            Reporter: Madhan Neethiraj
>            Assignee: Madhan Neethiraj
>
> For every 5 seconds following logs are added to /var/logs/atlas/application.log, which increases the log file size unnecessarily and makes the log difficult to use.
> {noformat}
> 2016-05-17 22:01:18,528 INFO - [pool-5-thread-1:] ~ Process identifier=hconnection-0x7878431f connecting to ZooKeeper ensemble=localhost:2181 (RecoverableZooKeeper:120)
> 2016-05-17 22:01:18,528 INFO - [pool-5-thread-1:] ~ Initiating client connection, connectString=localhost:2181 sessionTimeout=90000 watcher=org.apache.hadoop.hbase.zookeeper.PendingWatcher@78b60e6a (ZooKeeper:438)
> 2016-05-17 22:01:18,529 INFO - pool-5-thread-1-SendThread(localhost.localdomain:2181): ~ Opening socket connection to server localhost.localdomain/127.0.0.1:2181. Will not attempt to authenticate using SASL (unknown error) (ClientCnxn:1019)
> 2016-05-17 22:01:18,529 INFO - pool-5-thread-1-SendThread(localhost.localdomain:2181): ~ Socket connection established to localhost.localdomain/127.0.0.1:2181, initiating session (ClientCnxn:864)
> 2016-05-17 22:01:18,531 INFO - pool-5-thread-1-SendThread(localhost.localdomain:2181): ~ Session establishment complete on server localhost.localdomain/127.0.0.1:2181, sessionid = 0x154c1cf558e18b9, negotiated timeout = 40000 (ClientCnxn:1279)
> 2016-05-17 22:01:18,539 INFO - [pool-5-thread-1:] ~ Closing zookeeper sessionid=0x154c1cf558e18b9 (ConnectionManager$HConnectionImplementation:1684)
> 2016-05-17 22:01:18,540 INFO - [pool-5-thread-1:] ~ Session: 0x154c1cf558e18b9 closed (ZooKeeper:684)
> 2016-05-17 22:01:18,540 INFO - [pool-5-thread-1-EventThread:] ~ EventThread shut down (ClientCnxn:524)
> {noformat}



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)