[
https://issues.apache.org/jira/browse/HBASE-15205?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15149960#comment-15149960
]
stack commented on HBASE-15205:
-------------------------------
public NavigableMap<byte[], Integer> getScopes() { should be called
getReplicationScopes.
Does it have to be public? Is it only used inside in this class? It is odd
having this internal detail exported as public method on the Region Interface.
Sometimes you use getScopes. Other times you use this.replicationScopes. The
idea is that you get the scopes once on Region open? Thats great.
That write of a compaction marker that is in Store doesn't belong there
really.... (not your fault).
Should the replication scopes be in the WALKey and passed in that way when we
do an append (I didn't check so ignore if this is rubbish) Or, looking later,
this is where they end up anyways? In the WalKey? Do it when we create the
WALkey? Or scope is in the WALEntry and in the WALKey? Yeah, so you are adding
to WALEntry the scopes Map but it is in WALKey already?
Will we need HTableDescriptor down in WALEdit going forward? To get configs or
something from it? It is unused currently? Or in the append? Could you carry
the scopes in the HTD? The region puts them into the HTD key/value? The WAL
subsystem looks for them else falls back on old way of getting scopes? Thats a
bit hacky maybe.
You are breaking WALActionsListener? Need to release note that.
> Do not find the replication scope for every WAL#append()
> --------------------------------------------------------
>
> Key: HBASE-15205
> URL: https://issues.apache.org/jira/browse/HBASE-15205
> Project: HBase
> Issue Type: Sub-task
> Components: regionserver
> Reporter: ramkrishna.s.vasudevan
> Assignee: ramkrishna.s.vasudevan
> Priority: Minor
> Fix For: 2.0.0
>
> Attachments: HBASE-15205.patch, HBASE-15205_1.patch,
> HBASE-15205_2.patch, HBASE-15205_3.patch, ScopeWALEdits.jpg,
> ScopeWALEdits_afterpatch.jpg
>
>
> After the byte[] and char[] the other top contributor for lot of GC (though
> it is only 2.86%) is the UTF_8.newDecoder.
> This happens because for every WAL append we try to calculate the replication
> scope associate with the families associated with the TableDescriptor. I
> think per WAL append doing this is very costly and creates lot of garbage.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
