[ https://issues.apache.org/jira/browse/HDFS-10458?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15308994#comment-15308994 ]
Konstantin Shvachko commented on HDFS-10458: -------------------------------------------- The patch looks good. I would add a similar condition for empty {{encryptionZones}} at the start of {{EncryptionZoneManager.getEncryptionZoneForPath()}}. This method is used many times both for write operations, like {{startFile()}} and for reads, like {{getFileInfo()}}. Even though this will still be under the lock, but returning and releasing the lock quickly should be beneficial. > getFileEncryptionInfo should return quickly for non-encrypted cluster > --------------------------------------------------------------------- > > Key: HDFS-10458 > URL: https://issues.apache.org/jira/browse/HDFS-10458 > Project: Hadoop HDFS > Issue Type: Bug > Components: encryption, namenode > Affects Versions: 2.6.0 > Reporter: Zhe Zhang > Assignee: Zhe Zhang > Attachments: HDFS-10458.00.patch > > > {{FSDirectory#getFileEncryptionInfo}} always acquires {{readLock}} and checks > if the path belongs to an EZ. For a busy system with potentially many listing > operations, this could cause locking contention. > I think we should add a call {{EncryptionZoneManager#hasEncryptionZone()}} to > return whether the system has any EZ. If no EZ at all, > {{getFileEncryptionInfo}} should return null without {{readLock}}. > If {{hasEncryptionZone}} is only used in the above scenario, maybe itself > doesn't need a {{readLock}} -- if the system doesn't have any EZ when > {{getFileEncryptionInfo}} is called on a path, it means the path cannot be > encrypted. -- This message was sent by Atlassian JIRA (v6.3.4#6332) --------------------------------------------------------------------- To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org