[
https://issues.apache.org/jira/browse/MESOS-1332?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13993287#comment-13993287
]
Benjamin Mahler edited comment on MESOS-1332 at 5/9/14 5:57 PM:
----------------------------------------------------------------
(edit: removed "active_tasks" in Master's endpoint)
Here's what [~vinodkone], [~benjaminhindman], and myself came up with, taking
into account [~adam-mesos]'s comment
[here|https://issues.apache.org/jira/browse/MESOS-1188?focusedCommentId=13987547&page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#comment-13987547].
I've omitted the resource-related metrics as they can remain unchanged:
{code: title=Old Master's Metrics}
{
"elected": true,
"outstanding_offers": 124,
"uptime": 9938.216960768,
"activated_slaves": 127,
"deactivated_slaves": 0, // Removed slaves cache size (confusing).
"active_schedulers": 1,
"total_schedulers": 1, // This is active + inactive.
"failed_tasks": 647,
"finished_tasks": 5646,
"invalid_status_updates": 890,
"killed_tasks": 459,
"lost_tasks": 0,
"staged_tasks": 6728,
"started_tasks": 6087,
"active_tasks_gauge": 568,
"invalid_status_updates": 890,
"valid_status_updates": 18896
}
{code}
{code: title=New Master's Metrics}
{
"elected": true,
"outstanding_offers": 124,
"uptime_secs": 9938.216960768, // Adding "_secs".
"active_slaves": 127, // Now in present tense.
"inactive_slaves": 0, // Gauge of inactive (disconnected) slaves.
"active_frameworks": 1, // Change from "schedulers" to "frameworks".
"inactive_frameworks": 0, // Gauge of inactive frameworks instead of "total".
"tasks_staging": x, // Present tense, gauge instead.
"tasks_starting": x, // Present tense, gauge instead.
"tasks_running": x, // Present tense, gauge instead.
"tasks_failed": 647,
"tasks_finished": 5646,
"tasks_killed": 459,
"tasks_lost": 0,
"invalid_status_updates": 890,
"valid_status_updates": 18896
}
{code}
Now for the slave:
{code: title=Old Slave's Metrics}
{
"invalid_status_updates": 0,
"valid_status_updates": 170
"uptime": 14651.206068992,
"registered": "1",
"recovery_errors": 0,
"staged_tasks": 114,
"started_tasks": 50,
"failed_tasks": 4,
"finished_tasks": 52,
"killed_tasks": 4,
"lost_tasks": 0,
"launched_tasks_gauge": 3,
"queued_tasks_gauge": 0,
"total_frameworks": 1,
}
{code}
{code: title=New Slave's Metrics}
{
"invalid_status_updates": 0,
"valid_status_updates": 170
"uptime_secs": 14651.206068992, // Adding "_secs" unit.
"registered": "1",
"recovery_errors": 0,
"tasks_staging": x, // Gauge, this also obviates queued tasks.
"tasks_starting": x, // Gauge.
"tasks_running": x, // Gauge.
"tasks_failed": 4,
"tasks_finished": 52,
"tasks_killed": 4,
"tasks_lost": 0,
"active_frameworks": 1, // Now "active".
}
{code}
was (Author: bmahler):
Here's what [~vinodkone], [~benjaminhindman], and myself came up with, taking
into account [~adam-mesos]'s comment
[here|https://issues.apache.org/jira/browse/MESOS-1188?focusedCommentId=13987547&page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#comment-13987547].
I've omitted the resource-related metrics as they can remain unchanged:
{code: title=Old Master's Metrics}
{
"elected": true,
"outstanding_offers": 124,
"uptime": 9938.216960768,
"activated_slaves": 127,
"deactivated_slaves": 0, // Removed slaves cache size (confusing).
"active_schedulers": 1,
"total_schedulers": 1, // This is active + inactive.
"failed_tasks": 647,
"finished_tasks": 5646,
"invalid_status_updates": 890,
"killed_tasks": 459,
"lost_tasks": 0,
"staged_tasks": 6728,
"started_tasks": 6087,
"active_tasks_gauge": 568,
"invalid_status_updates": 890,
"valid_status_updates": 18896
}
{code}
{code: title=New Master's Metrics}
{
"elected": true,
"outstanding_offers": 124,
"uptime_secs": 9938.216960768, // Adding "_secs".
"active_slaves": 127, // Now in present tense.
"inactive_slaves": 0, // Gauge of inactive (disconnected) slaves.
"active_frameworks": 1, // Change from "schedulers" to "frameworks".
"inactive_frameworks": 0, // Gauge of inactive frameworks instead of "total".
"tasks_staging": x, // Present tense, gauge instead.
"tasks_starting": x, // Present tense, gauge instead.
"tasks_running": x, // Present tense, gauge instead.
"tasks_failed": 647,
"tasks_finished": 5646,
"tasks_killed": 459,
"tasks_lost": 0,
"active_tasks": 568,
"invalid_status_updates": 890,
"valid_status_updates": 18896
}
{code}
Now for the slave:
{code: title=Old Slave's Metrics}
{
"invalid_status_updates": 0,
"valid_status_updates": 170
"uptime": 14651.206068992,
"registered": "1",
"recovery_errors": 0,
"staged_tasks": 114,
"started_tasks": 50,
"failed_tasks": 4,
"finished_tasks": 52,
"killed_tasks": 4,
"lost_tasks": 0,
"launched_tasks_gauge": 3,
"queued_tasks_gauge": 0,
"total_frameworks": 1,
}
{code}
{code: title=New Slave's Metrics}
{
"invalid_status_updates": 0,
"valid_status_updates": 170
"uptime_secs": 14651.206068992, // Adding "_secs" unit.
"registered": "1",
"recovery_errors": 0,
"tasks_staging": x, // Gauge, this also obviates queued tasks.
"tasks_starting": x, // Gauge.
"tasks_running": x, // Gauge.
"tasks_failed": 4,
"tasks_finished": 52,
"tasks_killed": 4,
"tasks_lost": 0,
"active_frameworks": 1, // Now "active".
}
{code}
> Improve Master and Slave metric names
> -------------------------------------
>
> Key: MESOS-1332
> URL: https://issues.apache.org/jira/browse/MESOS-1332
> Project: Mesos
> Issue Type: Improvement
> Components: master, slave
> Reporter: Dominic Hamon
> Assignee: Dominic Hamon
>
> As we move the metrics to a new endpoint, we should consider revisiting the
> names of some of the current metrics to make them clearer.
> It may also be worth considering changing some existing counter-style metrics
> to gauges.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
