[
https://issues.apache.org/jira/browse/HDFS-2812?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13189437#comment-13189437
]
Todd Lipcon commented on HDFS-2812:
-----------------------------------
I accidentally used the {{haEnabled}} member variable in this patch, but that
was from an earlier patch I was building on top of which isn't committed yet.
The patch needs the following delta:
{code}
--- src/main/java/org/apache/hadoop/hdfs/server/namenode/FSNamesystem.java
(revision 1233595)
+++ src/main/java/org/apache/hadoop/hdfs/server/namenode/FSNamesystem.java
(working copy)
@@ -337,6 +337,8 @@
*/
private HAContext haContext;
+ private boolean haEnabled;
+
private final Configuration conf;
@@ -737,8 +746,8 @@
// block allocation has to be persisted in HA using a shared edits
directory
// so that the standby has up-to-date namespace information
String nameserviceId = DFSUtil.getNamenodeNameServiceId(conf);
- this.persistBlocks |= HAUtil.isHAEnabled(conf, nameserviceId) &&
- HAUtil.usesSharedEditsDir(conf);
+ this.haEnabled = HAUtil.isHAEnabled(conf, nameserviceId);
+ this.persistBlocks |= haEnabled && HAUtil.usesSharedEditsDir(conf);
short filePermission =
(short)conf.getInt(DFS_NAMENODE_UPGRADE_PERMISSION_KEY,
DFS_NAMENODE_UPGRADE_PERMISSION_DEFAULT);
{code}
> When becoming active, NN should treat all leases as freshly renewed
> -------------------------------------------------------------------
>
> Key: HDFS-2812
> URL: https://issues.apache.org/jira/browse/HDFS-2812
> Project: Hadoop HDFS
> Issue Type: Sub-task
> Components: ha, name-node
> Affects Versions: HA branch (HDFS-1623)
> Reporter: Todd Lipcon
> Assignee: Todd Lipcon
> Priority: Critical
> Attachments: hdfs-2812.txt, hdfs-2812.txt
>
>
> I haven't seen this yet in practice, but I think the following is a bug:
> - Clients currently only renew leases to the active NN
> - Leases are written into the log when created, but there is no record to
> renew them
> - In Standby state, we don't check leases for expiration. But, when we fail
> over and start the lease monitor, since the leases haven't been renewed, they
> might end up getting incorrectly released.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators:
https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira