You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@cassandra.apache.org by "Sammy Yu (JIRA)" <ji...@apache.org> on 2009/10/08 22:49:31 UTC

[jira] Created: (CASSANDRA-482) Expose out additional metrics for the thread pools: active threads, pool size, and completed tasks

Expose out additional metrics for the thread pools: active threads, pool size, and completed tasks
--------------------------------------------------------------------------------------------------

                 Key: CASSANDRA-482
                 URL: https://issues.apache.org/jira/browse/CASSANDRA-482
             Project: Cassandra
          Issue Type: New Feature
    Affects Versions: 0.5
            Reporter: Sammy Yu
            Assignee: Sammy Yu


It would be nice to expose out additional metrics in the threadpools via nodeprobe's tpstats to understand how busy the system is.



-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.


[jira] Commented: (CASSANDRA-482) Expose out additional metrics for the thread pools: active threads, pool size, and completed tasks

Posted by "Jonathan Ellis (JIRA)" <ji...@apache.org>.
    [ https://issues.apache.org/jira/browse/CASSANDRA-482?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12763740#action_12763740 ] 

Jonathan Ellis commented on CASSANDRA-482:
------------------------------------------

from irc:

> i can see runningtasks being useful actually, to see how close you are to an executor getting swamped, and completedtasks, so rrd can turn it into a rate which is useful for system health.  but poolsize is just not useful.

let's do this.

rename getRunningTasks to getActiveCount so it just leverages the TPE version.  make the CLE version always return 1.  (not strictly accurate, but it's basically useless info for a non-pooled executor.)

make the completed count in CLE a volatile long instead of atomic -- you only need an Atomic variable if multiple threads are going to update it which is not the case here.

> Expose out additional metrics for the thread pools: active threads, pool size, and completed tasks
> --------------------------------------------------------------------------------------------------
>
>                 Key: CASSANDRA-482
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-482
>             Project: Cassandra
>          Issue Type: New Feature
>    Affects Versions: 0.5
>            Reporter: Sammy Yu
>            Assignee: Sammy Yu
>         Attachments: 0001--CASSANDRA-482-Expose-out-additional-metrics-for-th.patch
>
>
> It would be nice to expose out additional metrics in the threadpools via nodeprobe's tpstats to understand how busy the system is.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.


[jira] Commented: (CASSANDRA-482) Expose out additional metrics for the thread pools: active threads, pool size, and completed tasks

Posted by "Hudson (JIRA)" <ji...@apache.org>.
    [ https://issues.apache.org/jira/browse/CASSANDRA-482?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12764339#action_12764339 ] 

Hudson commented on CASSANDRA-482:
----------------------------------

Integrated in Cassandra #223 (See [http://hudson.zones.apache.org/hudson/job/Cassandra/223/])
    expose additional metrics for executors.  patch by Sammy Yu; reviewed by jbellis for 


> Expose out additional metrics for the thread pools: active threads, pool size, and completed tasks
> --------------------------------------------------------------------------------------------------
>
>                 Key: CASSANDRA-482
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-482
>             Project: Cassandra
>          Issue Type: New Feature
>    Affects Versions: 0.5
>            Reporter: Sammy Yu
>            Assignee: Sammy Yu
>             Fix For: 0.5
>
>         Attachments: 0001--CASSANDRA-482-Expose-out-additional-metrics-for-th.patch, 0002--CASSANDRA-482-Expose-out-additional-metrics-for-th.patch
>
>
> It would be nice to expose out additional metrics in the threadpools via nodeprobe's tpstats to understand how busy the system is.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.


[jira] Updated: (CASSANDRA-482) Expose out additional metrics for the thread pools: active threads, pool size, and completed tasks

Posted by "Sammy Yu (JIRA)" <ji...@apache.org>.
     [ https://issues.apache.org/jira/browse/CASSANDRA-482?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Sammy Yu updated CASSANDRA-482:
-------------------------------

    Attachment: 0001--CASSANDRA-482-Expose-out-additional-metrics-for-th.patch

The new output of tpstats (disregard the spacing):
Pool Name                    Active      Size   Pending      Completed
MESSAGING-SERVICE-POOL            0         4         0              0
FILEUTILS-DELETE-POOL             0         1         0              0
RESPONSE-STAGE                    0         4         0              0
BOOT-STRAPPER                     0         1         0              0
ROW-READ-STAGE                    0         8         0              0
LB-OPERATIONS                     0         1         0              0
COMMITLOG                         1         1         0            344
GMFD                              0         1         0              0
MESSAGE-DESERIALIZER-POOL         0         4         0              0
LB-TARGET                         0         1         0              0
CONSISTENCY-MANAGER               0         4         0              0
ROW-MUTATION-STAGE                0        32         0              0
MESSAGE-STREAMING-POOL            0         1         0              0
LOAD-BALANCER-STAGE               0         1         0              0
FLUSH-SORTER-POOL                 0         1         0              0
MEMTABLE-POST-FLUSHER             0         1         0              0
COMPACTION-POOL                   0         1         0              4
FLUSH-WRITER-POOL                 0         2         0              0
HINTED-HANDOFF-POOL               0         1         0              0


The CommitLogExecutorService mapping hacks probably need to be looked at.


> Expose out additional metrics for the thread pools: active threads, pool size, and completed tasks
> --------------------------------------------------------------------------------------------------
>
>                 Key: CASSANDRA-482
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-482
>             Project: Cassandra
>          Issue Type: New Feature
>    Affects Versions: 0.5
>            Reporter: Sammy Yu
>            Assignee: Sammy Yu
>         Attachments: 0001--CASSANDRA-482-Expose-out-additional-metrics-for-th.patch
>
>
> It would be nice to expose out additional metrics in the threadpools via nodeprobe's tpstats to understand how busy the system is.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.


[jira] Updated: (CASSANDRA-482) Expose out additional metrics for the thread pools: active threads, pool size, and completed tasks

Posted by "Sammy Yu (JIRA)" <ji...@apache.org>.
     [ https://issues.apache.org/jira/browse/CASSANDRA-482?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Sammy Yu updated CASSANDRA-482:
-------------------------------

    Attachment: 0002--CASSANDRA-482-Expose-out-additional-metrics-for-th.patch

Incorporated jbellis' comments.

> Expose out additional metrics for the thread pools: active threads, pool size, and completed tasks
> --------------------------------------------------------------------------------------------------
>
>                 Key: CASSANDRA-482
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-482
>             Project: Cassandra
>          Issue Type: New Feature
>    Affects Versions: 0.5
>            Reporter: Sammy Yu
>            Assignee: Sammy Yu
>             Fix For: 0.5
>
>         Attachments: 0001--CASSANDRA-482-Expose-out-additional-metrics-for-th.patch, 0002--CASSANDRA-482-Expose-out-additional-metrics-for-th.patch
>
>
> It would be nice to expose out additional metrics in the threadpools via nodeprobe's tpstats to understand how busy the system is.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.


[jira] Updated: (CASSANDRA-482) Expose out additional metrics for the thread pools: active threads, pool size, and completed tasks

Posted by "Sammy Yu (JIRA)" <ji...@apache.org>.
     [ https://issues.apache.org/jira/browse/CASSANDRA-482?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Sammy Yu updated CASSANDRA-482:
-------------------------------

    Fix Version/s: 0.5

> Expose out additional metrics for the thread pools: active threads, pool size, and completed tasks
> --------------------------------------------------------------------------------------------------
>
>                 Key: CASSANDRA-482
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-482
>             Project: Cassandra
>          Issue Type: New Feature
>    Affects Versions: 0.5
>            Reporter: Sammy Yu
>            Assignee: Sammy Yu
>             Fix For: 0.5
>
>         Attachments: 0001--CASSANDRA-482-Expose-out-additional-metrics-for-th.patch
>
>
> It would be nice to expose out additional metrics in the threadpools via nodeprobe's tpstats to understand how busy the system is.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.


[jira] Commented: (CASSANDRA-482) Expose out additional metrics for the thread pools: active threads, pool size, and completed tasks

Posted by "Jonathan Ellis (JIRA)" <ji...@apache.org>.
    [ https://issues.apache.org/jira/browse/CASSANDRA-482?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12763730#action_12763730 ] 

Jonathan Ellis commented on CASSANDRA-482:
------------------------------------------

+    public int getRunningTasks();

This is not worth the overhead in the commitlog executor.

+    public int getPoolSize();

+    public long getCompletedTasks();

What actionable information do these produce?  I.e., what decision can you now make with this data that you could not before?

> Expose out additional metrics for the thread pools: active threads, pool size, and completed tasks
> --------------------------------------------------------------------------------------------------
>
>                 Key: CASSANDRA-482
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-482
>             Project: Cassandra
>          Issue Type: New Feature
>    Affects Versions: 0.5
>            Reporter: Sammy Yu
>            Assignee: Sammy Yu
>         Attachments: 0001--CASSANDRA-482-Expose-out-additional-metrics-for-th.patch
>
>
> It would be nice to expose out additional metrics in the threadpools via nodeprobe's tpstats to understand how busy the system is.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.