You are viewing a plain text version of this content. The canonical link for it is here.

Posted to issues@hbase.apache.org by "Ashu Pachauri (JIRA)" <ji...@apache.org> on 2017/04/08 00:08:41 UTC

[jira] [Created] (HBASE-17892) Hierarchical paths for WALs

Ashu Pachauri created HBASE-17892:
-------------------------------------

             Summary: Hierarchical paths for WALs
                 Key: HBASE-17892
                 URL: https://issues.apache.org/jira/browse/HBASE-17892
             Project: HBase
          Issue Type: Improvement
          Components: Replication, wal
            Reporter: Ashu Pachauri


Currently all WALs are just dumped into a single directory irrespective of the ServerName for the regionserver to which they belong.
This is a problem specifically for the WALs archive directory (HConstants.HREGION_OLDLOGDIR_NAME) since it holds a large number of files which can easily make it go above the HDFS max items in a directory limit for large clusters in scenarios like pausing replication. HDFS imposes a hard limit of around 6 million files in a single directory due to Protobuf serialization limitations.

We should put WALs for a particular servername in its own directory inside the WALs (archive) directory.



--
This message was sent by Atlassian JIRA
(v6.3.15#6346)