[jira] [Commented] (CASSANDRA-6423) Some histogram metrics of long[] type are unusable with Graphite

2013-12-04 Thread mat gomes (JIRA)

[ 
https://issues.apache.org/jira/browse/CASSANDRA-6423?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13839219#comment-13839219
 ] 

mat gomes commented on CASSANDRA-6423:
--

We are also having the same issue.
invalid line received from client IP ignoring. Flooding the logs

Please share resolutions to this issue.
Thanks.


 Some histogram metrics of long[] type are unusable with Graphite
 

 Key: CASSANDRA-6423
 URL: https://issues.apache.org/jira/browse/CASSANDRA-6423
 Project: Cassandra
  Issue Type: Bug
  Components: Core
 Environment: Cassandra 2.0.3, Oracle Linux 6.3, graphite
Reporter: Nikolai Grigoriev
Priority: Minor

 I am not entirely sure if this is a Cassandra issue or the limitation of the 
 graphite reporter agent. But since the metrics in question are created by 
 Cassandra itself I have decided that it may be appropriate to report it here.
 I am using Graphite with Cassandra 2.0.x and I have recently noticed frequent 
 'invalid line' messages in Graphite log. Unfortunately Graphite did not 
 provide enough details so I hacked it a bit to get more verbose error message 
 with the metric name. And here is what I have found:
 {code}
 29/11/2013 16:07:16 :: invalid line 
 .org.apache.cassandra.metrics.ColumnFamily.mykeyspace.some_cf.EstimatedRowSizeHistogram.value
  [J@17c26ba5 1385741236received from client 192.168.20.157:3, ignoring
 29/11/2013 16:07:16 :: invalid line 
 .org.apache.cassandra.metrics.ColumnFamily.mykeyspace.some_cf.EstimatedColumnCountHistogram.value
  [J@5d2929d2 1385741236received from client 192.168.20.157:3, ignoring
 29/11/2013 16:07:16 :: invalid line 
 .org.apache.cassandra.metrics.ColumnFamily.mykeyspace.some_cf.EstimatedRowSizeHistogram.value
  [J@3978c9c6 1385741236received from client 192.168.20.157:3, ignoring
 29/11/2013 16:07:16 :: invalid line 
 .org.apache.cassandra.metrics.ColumnFamily.mykeyspace.some_cf.EstimatedColumnCountHistogram.value
  [J@290703a4 1385741236received from client 192.168.20.157:3, ignoring
 29/11/2013 16:07:16 :: invalid line 
 .org.apache.cassandra.metrics.ColumnFamily.mykeyspace.some_cf.EstimatedRowSizeHistogram.value
  [J@b801907 1385741236received from client 192.168.20.157:3, ignoring
 {code}
 Then a quick search through Cassandre code confirmed that there is a number 
 of histograms created as Gauge with long[] data. So, when they are serialized 
 by  GraphiteReporter they are just printed as long[].toString(), making these 
 metrics useless.
 I am not sure what would be the best solution to it. I do see some histograms 
 (LiveScannedHistogram etc) that are implemented differently and they are 
 properly sent to Graphite.



--
This message was sent by Atlassian JIRA
(v6.1#6144)


[jira] [Commented] (CASSANDRA-6423) Some histogram metrics of long[] type are unusable with Graphite

2013-12-04 Thread Mikhail Stepura (JIRA)

[ 
https://issues.apache.org/jira/browse/CASSANDRA-6423?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13839449#comment-13839449
 ] 

Mikhail Stepura commented on CASSANDRA-6423:


I wonder if it's possible to convert those {{Gaugelong[]}} to {{Histogram}}, 
i.e. populate metric's  {{Histogram}} from  {{EstimatedHistogram}}

 Some histogram metrics of long[] type are unusable with Graphite
 

 Key: CASSANDRA-6423
 URL: https://issues.apache.org/jira/browse/CASSANDRA-6423
 Project: Cassandra
  Issue Type: Bug
  Components: Core
 Environment: Cassandra 2.0.3, Oracle Linux 6.3, graphite
Reporter: Nikolai Grigoriev
Priority: Minor

 I am not entirely sure if this is a Cassandra issue or the limitation of the 
 graphite reporter agent. But since the metrics in question are created by 
 Cassandra itself I have decided that it may be appropriate to report it here.
 I am using Graphite with Cassandra 2.0.x and I have recently noticed frequent 
 'invalid line' messages in Graphite log. Unfortunately Graphite did not 
 provide enough details so I hacked it a bit to get more verbose error message 
 with the metric name. And here is what I have found:
 {code}
 29/11/2013 16:07:16 :: invalid line 
 .org.apache.cassandra.metrics.ColumnFamily.mykeyspace.some_cf.EstimatedRowSizeHistogram.value
  [J@17c26ba5 1385741236received from client 192.168.20.157:3, ignoring
 29/11/2013 16:07:16 :: invalid line 
 .org.apache.cassandra.metrics.ColumnFamily.mykeyspace.some_cf.EstimatedColumnCountHistogram.value
  [J@5d2929d2 1385741236received from client 192.168.20.157:3, ignoring
 29/11/2013 16:07:16 :: invalid line 
 .org.apache.cassandra.metrics.ColumnFamily.mykeyspace.some_cf.EstimatedRowSizeHistogram.value
  [J@3978c9c6 1385741236received from client 192.168.20.157:3, ignoring
 29/11/2013 16:07:16 :: invalid line 
 .org.apache.cassandra.metrics.ColumnFamily.mykeyspace.some_cf.EstimatedColumnCountHistogram.value
  [J@290703a4 1385741236received from client 192.168.20.157:3, ignoring
 29/11/2013 16:07:16 :: invalid line 
 .org.apache.cassandra.metrics.ColumnFamily.mykeyspace.some_cf.EstimatedRowSizeHistogram.value
  [J@b801907 1385741236received from client 192.168.20.157:3, ignoring
 {code}
 Then a quick search through Cassandre code confirmed that there is a number 
 of histograms created as Gauge with long[] data. So, when they are serialized 
 by  GraphiteReporter they are just printed as long[].toString(), making these 
 metrics useless.
 I am not sure what would be the best solution to it. I do see some histograms 
 (LiveScannedHistogram etc) that are implemented differently and they are 
 properly sent to Graphite.



--
This message was sent by Atlassian JIRA
(v6.1#6144)


[jira] [Commented] (CASSANDRA-6423) Some histogram metrics of long[] type are unusable with Graphite

2013-11-29 Thread Nikolai Grigoriev (JIRA)

[ 
https://issues.apache.org/jira/browse/CASSANDRA-6423?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13835475#comment-13835475
 ] 

Nikolai Grigoriev commented on CASSANDRA-6423:
--

I think the consumers cannot really be expected to understand 
[J@290703a4...Knowing how the serialization of the metrics works, can you 
really expect that your metric defined as long[] would be serialized as 
something meaningful?

 Some histogram metrics of long[] type are unusable with Graphite
 

 Key: CASSANDRA-6423
 URL: https://issues.apache.org/jira/browse/CASSANDRA-6423
 Project: Cassandra
  Issue Type: Bug
  Components: Core
 Environment: Cassandra 2.0.3, Oracle Linux 6.3, graphite
Reporter: Nikolai Grigoriev
Priority: Minor

 I am not entirely sure if this is a Cassandra issue or the limitation of the 
 graphite reporter agent. But since the metrics in question are created by 
 Cassandra itself I have decided that it may be appropriate to report it here.
 I am using Graphite with Cassandra 2.0.x and I have recently noticed frequent 
 'invalid line' messages in Graphite log. Unfortunately Graphite did not 
 provide enough details so I hacked it a bit to get more verbose error message 
 with the metric name. And here is what I have found:
 {code}
 29/11/2013 16:07:16 :: invalid line 
 .org.apache.cassandra.metrics.ColumnFamily.mykeyspace.some_cf.EstimatedRowSizeHistogram.value
  [J@17c26ba5 1385741236received from client 192.168.20.157:3, ignoring
 29/11/2013 16:07:16 :: invalid line 
 .org.apache.cassandra.metrics.ColumnFamily.mykeyspace.some_cf.EstimatedColumnCountHistogram.value
  [J@5d2929d2 1385741236received from client 192.168.20.157:3, ignoring
 29/11/2013 16:07:16 :: invalid line 
 .org.apache.cassandra.metrics.ColumnFamily.mykeyspace.some_cf.EstimatedRowSizeHistogram.value
  [J@3978c9c6 1385741236received from client 192.168.20.157:3, ignoring
 29/11/2013 16:07:16 :: invalid line 
 .org.apache.cassandra.metrics.ColumnFamily.mykeyspace.some_cf.EstimatedColumnCountHistogram.value
  [J@290703a4 1385741236received from client 192.168.20.157:3, ignoring
 29/11/2013 16:07:16 :: invalid line 
 .org.apache.cassandra.metrics.ColumnFamily.mykeyspace.some_cf.EstimatedRowSizeHistogram.value
  [J@b801907 1385741236received from client 192.168.20.157:3, ignoring
 {code}
 Then a quick search through Cassandre code confirmed that there is a number 
 of histograms created as Gauge with long[] data. So, when they are serialized 
 by  GraphiteReporter they are just printed as long[].toString(), making these 
 metrics useless.
 I am not sure what would be the best solution to it. I do see some histograms 
 (LiveScannedHistogram etc) that are implemented differently and they are 
 properly sent to Graphite.



--
This message was sent by Atlassian JIRA
(v6.1#6144)


[jira] [Commented] (CASSANDRA-6423) Some histogram metrics of long[] type are unusable with Graphite

2013-11-29 Thread Mikhail Stepura (JIRA)

[ 
https://issues.apache.org/jira/browse/CASSANDRA-6423?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13835552#comment-13835552
 ] 

Mikhail Stepura commented on CASSANDRA-6423:


https://github.com/codahale/metrics/issues is a good place to start this 
discussion.
At the moment GraphiteReporter doesn't handle Gauges which return an array 
(which is perfectly valid use case) instead of a single value.
I think checking {{object.getClass().isArray()}} and using 
{{Arrays.toString(object)}} should help

 Some histogram metrics of long[] type are unusable with Graphite
 

 Key: CASSANDRA-6423
 URL: https://issues.apache.org/jira/browse/CASSANDRA-6423
 Project: Cassandra
  Issue Type: Bug
  Components: Core
 Environment: Cassandra 2.0.3, Oracle Linux 6.3, graphite
Reporter: Nikolai Grigoriev
Priority: Minor

 I am not entirely sure if this is a Cassandra issue or the limitation of the 
 graphite reporter agent. But since the metrics in question are created by 
 Cassandra itself I have decided that it may be appropriate to report it here.
 I am using Graphite with Cassandra 2.0.x and I have recently noticed frequent 
 'invalid line' messages in Graphite log. Unfortunately Graphite did not 
 provide enough details so I hacked it a bit to get more verbose error message 
 with the metric name. And here is what I have found:
 {code}
 29/11/2013 16:07:16 :: invalid line 
 .org.apache.cassandra.metrics.ColumnFamily.mykeyspace.some_cf.EstimatedRowSizeHistogram.value
  [J@17c26ba5 1385741236received from client 192.168.20.157:3, ignoring
 29/11/2013 16:07:16 :: invalid line 
 .org.apache.cassandra.metrics.ColumnFamily.mykeyspace.some_cf.EstimatedColumnCountHistogram.value
  [J@5d2929d2 1385741236received from client 192.168.20.157:3, ignoring
 29/11/2013 16:07:16 :: invalid line 
 .org.apache.cassandra.metrics.ColumnFamily.mykeyspace.some_cf.EstimatedRowSizeHistogram.value
  [J@3978c9c6 1385741236received from client 192.168.20.157:3, ignoring
 29/11/2013 16:07:16 :: invalid line 
 .org.apache.cassandra.metrics.ColumnFamily.mykeyspace.some_cf.EstimatedColumnCountHistogram.value
  [J@290703a4 1385741236received from client 192.168.20.157:3, ignoring
 29/11/2013 16:07:16 :: invalid line 
 .org.apache.cassandra.metrics.ColumnFamily.mykeyspace.some_cf.EstimatedRowSizeHistogram.value
  [J@b801907 1385741236received from client 192.168.20.157:3, ignoring
 {code}
 Then a quick search through Cassandre code confirmed that there is a number 
 of histograms created as Gauge with long[] data. So, when they are serialized 
 by  GraphiteReporter they are just printed as long[].toString(), making these 
 metrics useless.
 I am not sure what would be the best solution to it. I do see some histograms 
 (LiveScannedHistogram etc) that are implemented differently and they are 
 properly sent to Graphite.



--
This message was sent by Atlassian JIRA
(v6.1#6144)