[
https://issues.apache.org/jira/browse/HBASE-16768?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15552674#comment-15552674
]
Ted Yu commented on HBASE-16768:
--------------------------------
{code}
638 public long add(final Iterable<Cell> cells) {
639 lock.readLock().lock();
{code}
Why is read lock taken instead of the write lock ?
> Inconsistent results from the Append/Increment
> ----------------------------------------------
>
> Key: HBASE-16768
> URL: https://issues.apache.org/jira/browse/HBASE-16768
> Project: HBase
> Issue Type: Bug
> Reporter: ChiaPing Tsai
> Assignee: ChiaPing Tsai
> Priority: Minor
> Fix For: 2.0.0
>
> Attachments: HBASE-16768.v0.patch, HBASE-16768.v1.patch,
> HBASE-16768.v2.patch
>
>
> If the Append/Increment skips the WAL, the sequential id will not be updated.
> It introduces that the AbstractMemStore#upsert may remove the cell even
> though the cell is visible to the oldest scanner.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
