[
https://issues.apache.org/jira/browse/HDFS-158?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Aaron T. Myers resolved HDFS-158.
---------------------------------
Resolution: Fixed
This was long ago fixed through the implementation of sync and the recoverLease
API.
> HDFS soft lease limit should be configurable per-file
> -----------------------------------------------------
>
> Key: HDFS-158
> URL: https://issues.apache.org/jira/browse/HDFS-158
> Project: Hadoop HDFS
> Issue Type: Bug
> Environment: All operating systems
> Reporter: Doug Judd
>
> In Hypertable, there is a process called the RangeServer which is responsible
> for handling updates and queries on portions (ranges) of a database table.
> When an update arrives at the RangeServer, it first gets written to a commit
> log and then it goes into an in-memory table structure. If at any time, the
> RangeServer dies, when it comes back up, it replays the commit log to
> reconstruct the state of the in-memory table structure. As it stands now,
> when the RangeServer comes up, it cannot read the commit log file until the
> soft lease limit has expired. This causes the RangeServer to hang with all
> of the ranges that it was managing unaccessible.
> We would like to be able to configure the soft lease limit per-file so that
> we can set the limit to 0 for our commit log files. That would allow
> Hypertable to recover immediately after a crash.
--
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
