hbase-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Heng Chen (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HBASE-15740) Replication source.shippedKBs metric is undercounting because it is in KB
Date Sat, 30 Apr 2016 00:47:15 GMT

    [ https://issues.apache.org/jira/browse/HBASE-15740?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15265019#comment-15265019
] 

Heng Chen commented on HBASE-15740:
-----------------------------------

{code}
-  @Override public void incrShippedKBs(long size) {
+  @Override public void incrShippedBytes(long size) {
     shippedKBsCounter.incr(size);
+    MetricsReplicationGlobalSourceSource
+      .incrementKBsCounter(shippedBytesCounter, shippedKBsCounter);
   }
{code}

shippedKBsCounter -> shippedBytesCounter  ?

> Replication source.shippedKBs metric is undercounting because it is in KB
> -------------------------------------------------------------------------
>
>                 Key: HBASE-15740
>                 URL: https://issues.apache.org/jira/browse/HBASE-15740
>             Project: HBase
>          Issue Type: Sub-task
>            Reporter: Enis Soztutar
>            Assignee: Enis Soztutar
>             Fix For: 2.0.0, 1.3.0, 1.4.0
>
>         Attachments: hbase-15740_v1.patch
>
>
> In a cluster where there is replication going on, I've noticed that this is always 0:
> {code}
>     "source.shippedKBs" : 0,
> {code}
> Looking at the source reveals why:
> {code}
>           metrics.shipBatch(currentNbOperations, currentSize / 1024, currentNbHFiles);
> {code}
> It is always undercounting because we discard remaining bytes after KB boundary. This
is specially a problem when we are always shipping small batches <1KB.  



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Mime
View raw message