[
https://issues.apache.org/jira/browse/HBASE-20542?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16516158#comment-16516158
]
Ted Yu commented on HBASE-20542:
--------------------------------
Took a quick look.
{code}
+ while(!succ) {
+ currentActive = getActive();
+ succ = preUpdate(currentActive, cell, memstoreSizing);
{code}
Potentially how many {{preUpdate}} calls would take place when there is
contention ?
{code}
+ * @return true if the cell can be added to the
*/
@Override
- protected void checkActiveSize() {
- return;
+ protected boolean checkAndAddToActiveSize(MutableSegment currActive, Cell
cellToAdd,
{code}
The javadoc for @return is incomplete.
> Better heap utilization for IMC with MSLABs
> -------------------------------------------
>
> Key: HBASE-20542
> URL: https://issues.apache.org/jira/browse/HBASE-20542
> Project: HBase
> Issue Type: Sub-task
> Reporter: Eshcar Hillel
> Assignee: Eshcar Hillel
> Priority: Major
> Attachments: HBASE-20542.branch-2.001.patch
>
>
> Following HBASE-20188 we realized in-memory compaction combined with MSLABs
> may suffer from heap under-utilization due to internal fragmentation. This
> jira presents a solution to circumvent this problem. The main idea is to have
> each update operation check if it will cause overflow in the active segment
> *before* it is writing the new value (instead of checking the size after the
> write is completed), and if it is then the active segment is atomically
> swapped with a new empty segment, and is pushed (full-yet-not-overflowed) to
> the compaction pipeline. Later on the IMC deamon will run its compaction
> operation (flatten index/merge indices/data compaction) in the background.
> Some subtle concurrency issues should be handled with care. We next elaborate
> on them.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
