[jira] [Commented] (KAFKA-604) Add missing metrics in 0.8

2013-02-19 Thread Swapnil Ghike (JIRA)

[ 
https://issues.apache.org/jira/browse/KAFKA-604?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13581160#comment-13581160
 ] 

Swapnil Ghike commented on KAFKA-604:
-

I can do this after finishing KAFKA-755.

 Add missing metrics in 0.8
 --

 Key: KAFKA-604
 URL: https://issues.apache.org/jira/browse/KAFKA-604
 Project: Kafka
  Issue Type: Bug
  Components: core
Affects Versions: 0.8
Reporter: Jun Rao
Priority: Blocker
 Attachments: kafka_604_v1.patch, kafka_604_v2.patch, kafka_604_v3.diff

   Original Estimate: 24h
  Remaining Estimate: 24h

 It would be good if we add the following metrics:
 Producer: droppedMessageRate per topic
 ReplicaManager: partition count on the broker
 FileMessageSet: logFlushTimer per log (i.e., partition). Also, logFlushTime 
 should probably be moved to LogSegment since the flush now includes index 
 flush time.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira


[jira] [Commented] (KAFKA-604) Add missing metrics in 0.8

2013-02-18 Thread Jun Rao (JIRA)

[ 
https://issues.apache.org/jira/browse/KAFKA-604?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13580890#comment-13580890
 ] 

Jun Rao commented on KAFKA-604:
---

This patch no longer applies. Thinking a bit. We probably just need to get the 
following stats in jmx. The per topic level flush time is probably overkill.

Producer: droppedMessageRate per topic 

ReplicaManager: partition count on the broker 

 Add missing metrics in 0.8
 --

 Key: KAFKA-604
 URL: https://issues.apache.org/jira/browse/KAFKA-604
 Project: Kafka
  Issue Type: Bug
  Components: core
Affects Versions: 0.8
Reporter: Jun Rao
 Attachments: kafka_604_v1.patch, kafka_604_v2.patch, kafka_604_v3.diff

   Original Estimate: 24h
  Remaining Estimate: 24h

 It would be good if we add the following metrics:
 Producer: droppedMessageRate per topic
 ReplicaManager: partition count on the broker
 FileMessageSet: logFlushTimer per log (i.e., partition). Also, logFlushTime 
 should probably be moved to LogSegment since the flush now includes index 
 flush time.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira


[jira] [Commented] (KAFKA-604) Add missing metrics in 0.8

2013-02-18 Thread Jun Rao (JIRA)

[ 
https://issues.apache.org/jira/browse/KAFKA-604?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13580946#comment-13580946
 ] 

Jun Rao commented on KAFKA-604:
---

Also, in ProducerSendThread, we should get rid of the thread id part from the 
jmx bean name.
  newGauge(clientId + -ProducerQueueSize- + getId,
  new Gauge[Int] {
def getValue = queue.size
  })


 Add missing metrics in 0.8
 --

 Key: KAFKA-604
 URL: https://issues.apache.org/jira/browse/KAFKA-604
 Project: Kafka
  Issue Type: Bug
  Components: core
Affects Versions: 0.8
Reporter: Jun Rao
Priority: Blocker
 Attachments: kafka_604_v1.patch, kafka_604_v2.patch, kafka_604_v3.diff

   Original Estimate: 24h
  Remaining Estimate: 24h

 It would be good if we add the following metrics:
 Producer: droppedMessageRate per topic
 ReplicaManager: partition count on the broker
 FileMessageSet: logFlushTimer per log (i.e., partition). Also, logFlushTime 
 should probably be moved to LogSegment since the flush now includes index 
 flush time.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira


[jira] [Commented] (KAFKA-604) Add missing metrics in 0.8

2012-12-12 Thread Joel Koshy (JIRA)

[ 
https://issues.apache.org/jira/browse/KAFKA-604?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13530165#comment-13530165
 ] 

Joel Koshy commented on KAFKA-604:
--

The rebased patch does not really comment on/address the concern in the 
original review - i.e., it breaks a relatively clean
and convenient syntax in the common case (of updating a single timer) to 
support updates to multiple timers. It's obviously
not a particularly major issue, but a rung higher than say, a debate over 
whitespace/coding style. My only point is that we
should try and avoid making a change that goes from a nice syntax to a rather 
inconvenient syntax. For that reason, I'm
more in favor of the change to KafkaTimer in KAFKA-646.

That said, how about the following: both these patches need multiple timer 
updates to update an aggregate timer as well
as a specific timer - i.e., up to this point we don't really have a use case 
for updating more than two timers simultaneously.
So we can accomplish this case with the following:

aggregatekafkatimer.time {
  specifickafkatimer.time {
code block
  }
}

and avoid any change to KafkaTimer. Does that seem reasonable to you guys?


 Add missing metrics in 0.8
 --

 Key: KAFKA-604
 URL: https://issues.apache.org/jira/browse/KAFKA-604
 Project: Kafka
  Issue Type: Bug
  Components: core
Affects Versions: 0.8
Reporter: Jun Rao
 Attachments: kafka_604_v1.patch, kafka_604_v2.patch, kafka_604_v3.diff

   Original Estimate: 24h
  Remaining Estimate: 24h

 It would be good if we add the following metrics:
 Producer: droppedMessageRate per topic
 ReplicaManager: partition count on the broker
 FileMessageSet: logFlushTimer per log (i.e., partition). Also, logFlushTime 
 should probably be moved to LogSegment since the flush now includes index 
 flush time.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira


[jira] [Commented] (KAFKA-604) Add missing metrics in 0.8

2012-12-12 Thread Swapnil Ghike (JIRA)

[ 
https://issues.apache.org/jira/browse/KAFKA-604?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13530200#comment-13530200
 ] 

Swapnil Ghike commented on KAFKA-604:
-

Yes I like this idea, can follow this in KAFKA-646.

 Add missing metrics in 0.8
 --

 Key: KAFKA-604
 URL: https://issues.apache.org/jira/browse/KAFKA-604
 Project: Kafka
  Issue Type: Bug
  Components: core
Affects Versions: 0.8
Reporter: Jun Rao
 Attachments: kafka_604_v1.patch, kafka_604_v2.patch, kafka_604_v3.diff

   Original Estimate: 24h
  Remaining Estimate: 24h

 It would be good if we add the following metrics:
 Producer: droppedMessageRate per topic
 ReplicaManager: partition count on the broker
 FileMessageSet: logFlushTimer per log (i.e., partition). Also, logFlushTime 
 should probably be moved to LogSegment since the flush now includes index 
 flush time.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira


[jira] [Commented] (KAFKA-604) Add missing metrics in 0.8

2012-12-12 Thread Yang Ye (JIRA)

[ 
https://issues.apache.org/jira/browse/KAFKA-604?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13530223#comment-13530223
 ] 

Yang Ye commented on KAFKA-604:
---

Joel, thanks for the idea, it's good and I will do a new patch

Sent from my iPhone




 Add missing metrics in 0.8
 --

 Key: KAFKA-604
 URL: https://issues.apache.org/jira/browse/KAFKA-604
 Project: Kafka
  Issue Type: Bug
  Components: core
Affects Versions: 0.8
Reporter: Jun Rao
 Attachments: kafka_604_v1.patch, kafka_604_v2.patch, kafka_604_v3.diff

   Original Estimate: 24h
  Remaining Estimate: 24h

 It would be good if we add the following metrics:
 Producer: droppedMessageRate per topic
 ReplicaManager: partition count on the broker
 FileMessageSet: logFlushTimer per log (i.e., partition). Also, logFlushTime 
 should probably be moved to LogSegment since the flush now includes index 
 flush time.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira


[jira] [Commented] (KAFKA-604) Add missing metrics in 0.8

2012-12-11 Thread Swapnil Ghike (JIRA)

[ 
https://issues.apache.org/jira/browse/KAFKA-604?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13529626#comment-13529626
 ] 

Swapnil Ghike commented on KAFKA-604:
-

Wow, this is going to be fun. :) Unfortunately patch v3 here has a lot of 
conflicts with patch v5 at KAFKA-646 where I have done a wholesale 
reorganization of metrics. The conflicts are mostly in the use of KafkaTimer.

I agree that having a KafkaTimer object might be more efficient at runtime, but 
the tradeoff is that it changes the syntax of time measurement with some 
instances where we will need to pass in ({a bunch of statements spread out over 
multiple lines in a block}, timer).

Victor, can you take a look at patch v5 at the other jira and see which style 
of time measurement you prefer?

Perhaps Joel can also comment on the syntax changes. 

 Add missing metrics in 0.8
 --

 Key: KAFKA-604
 URL: https://issues.apache.org/jira/browse/KAFKA-604
 Project: Kafka
  Issue Type: Bug
  Components: core
Affects Versions: 0.8
Reporter: Jun Rao
 Attachments: kafka_604_v1.patch, kafka_604_v2.patch, kafka_604_v3.diff

   Original Estimate: 24h
  Remaining Estimate: 24h

 It would be good if we add the following metrics:
 Producer: droppedMessageRate per topic
 ReplicaManager: partition count on the broker
 FileMessageSet: logFlushTimer per log (i.e., partition). Also, logFlushTime 
 should probably be moved to LogSegment since the flush now includes index 
 flush time.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira


[jira] [Commented] (KAFKA-604) Add missing metrics in 0.8

2012-12-10 Thread Yang Ye (JIRA)

[ 
https://issues.apache.org/jira/browse/KAFKA-604?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13528116#comment-13528116
 ] 

Yang Ye commented on KAFKA-604:
---

Sure, I'll do that soon

Best,
---
Victor Yang Ye
+1(650)283-6547

http://www.linkedin.com/pub/victor-yang-ye/13/ba3/4b8http://www.linkedin.com/profile/view?id=47740172
http://www.facebook.com/yeyangever

Founder   Uyan.cc
Software Engineer in Distributed Data System Group, LinkedIn Corporation
Dept. of Computer Science, Graduate School of Art and Scient, Columbia
University
Special Pilot Computer Science Class, Tsinghua University

yeyange...@gmail.com
y...@linkedin.com
yy2...@columbia.edu






 Add missing metrics in 0.8
 --

 Key: KAFKA-604
 URL: https://issues.apache.org/jira/browse/KAFKA-604
 Project: Kafka
  Issue Type: Bug
  Components: core
Affects Versions: 0.8
Reporter: Jun Rao
 Attachments: kafka_604_v1.patch, kafka_604_v2.patch

   Original Estimate: 24h
  Remaining Estimate: 24h

 It would be good if we add the following metrics:
 Producer: droppedMessageRate per topic
 ReplicaManager: partition count on the broker
 FileMessageSet: logFlushTimer per log (i.e., partition). Also, logFlushTime 
 should probably be moved to LogSegment since the flush now includes index 
 flush time.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira