You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@flink.apache.org by GitBox <gi...@apache.org> on 2021/07/19 08:16:17 UTC
[GitHub] [flink] rkhachatryan commented on a change in pull request #16447: [FLINK-23144][metrics] Add percentiles to checkpoint statistics
rkhachatryan commented on a change in pull request #16447:
URL: https://github.com/apache/flink/pull/16447#discussion_r672086792
##########
File path: flink-runtime-web/web-dashboard/src/app/pages/job/checkpoints/job-checkpoints.component.html
##########
@@ -193,15 +193,74 @@
<td>{{ checkPointStats['summary']['state_size']['avg'] | humanizeBytes }}</td>
<td>{{ checkPointStats['summary']['processed_data']['avg'] | humanizeBytes }} ({{ checkPointStats['summary']['persisted_data']['avg'] | humanizeBytes }})</td>
</tr>
- <tr>
- <td><strong>Maximum</strong></td>
- <td>{{ checkPointStats['summary']['end_to_end_duration']['max'] | humanizeDuration}}</td>
- <td>{{ checkPointStats['summary']['state_size']['max'] | humanizeBytes }}</td>
- <td>{{ checkPointStats['summary']['processed_data']['max'] | humanizeBytes }} ({{ checkPointStats['summary']['persisted_data']['max'] | humanizeBytes }})</td>
- </tr>
+ <tr>
+ <td><strong>Maximum</strong></td>
+ <td>{{ checkPointStats['summary']['end_to_end_duration']['max'] | humanizeDuration}}</td>
+ <td>{{ checkPointStats['summary']['state_size']['max'] | humanizeBytes }}</td>
+ <td>{{ checkPointStats['summary']['processed_data']['max'] | humanizeBytes }} ({{ checkPointStats['summary']['persisted_data']['max'] | humanizeBytes }})</td>
+ </tr>
</ng-container>
</tbody>
</nz-table>
+
+ <nz-collapse>
+ <nz-collapse-panel
+ [nzHeader]="'Percentiles'"
+ [nzActive]="moreDetailsPanel.active"
+ [nzDisabled]="moreDetailsPanel.disabled">
+ <nz-table
+ *ngIf="checkPointStats"
+ class="no-border small"
+ [nzData]="(checkPointStats&&checkPointStats['summary'])?['']:[]"
+ [nzSize]="'small'"
+ [nzFrontPagination]="false"
+ [nzShowPagination]="false">
+ <thead>
+ <tr>
+ <th></th>
+ <th><strong>End to End Duration</strong></th>
+ <th><strong>Checkpointed Data Size</strong></th>
+ <th><strong>Processed (persisted) in-flight data</strong></th>
+ </tr>
+ </thead>
+ <tbody>
+ <ng-container *ngIf="checkPointStats['summary']">
+ <tr>
+ <td><strong>50% percentile</strong></td>
+ <td>{{ checkPointStats['summary']['end_to_end_duration']['p50'] | humanizeDuration}}</td>
+ <td>{{ checkPointStats['summary']['state_size']['p50'] | humanizeBytes }}</td>
+ <td>{{ checkPointStats['summary']['processed_data']['p50'] | humanizeBytes }} ({{ checkPointStats['summary']['persisted_data']['p50'] | humanizeBytes }})</td>
+ </tr>
+ <tr>
+ <td><strong>90% percentile</strong></td>
+ <td>{{ checkPointStats['summary']['end_to_end_duration']['p90'] | humanizeDuration}}</td>
+ <td>{{ checkPointStats['summary']['state_size']['p90'] | humanizeBytes }}</td>
+ <td>{{ checkPointStats['summary']['processed_data']['p90'] | humanizeBytes }} ({{ checkPointStats['summary']['persisted_data']['p90'] | humanizeBytes }})</td>
+ </tr>
+ <tr>
+ <td><strong>99% percentile</strong></td>
+ <td>{{ checkPointStats['summary']['end_to_end_duration']['p99'] | humanizeDuration}}</td>
+ <td>{{ checkPointStats['summary']['state_size']['p99'] | humanizeBytes }}</td>
+ <td>{{ checkPointStats['summary']['processed_data']['p99'] | humanizeBytes }} ({{ checkPointStats['summary']['persisted_data']['p99'] | humanizeBytes }})</td>
+ </tr>
+ <tr>
+ <td><strong>99% percentile</strong></td>
+ <td>{{ checkPointStats['summary']['end_to_end_duration']['p99'] | humanizeDuration}}</td>
+ <td>{{ checkPointStats['summary']['state_size']['p99'] | humanizeBytes }}</td>
+ <td>{{ checkPointStats['summary']['processed_data']['p99'] | humanizeBytes }} ({{ checkPointStats['summary']['persisted_data']['p99'] | humanizeBytes }})</td>
+ </tr>
+ <tr>
+ <td><strong>99.9% percentile</strong></td>
Review comment:
Thanks for reviewing!
The aggregation here is over time, not over tasks. Each data point is a checkpoint, without details per-subtask (subtask details are aggregated upon finalization using sum/max).
And the "time" is currently limited to 10K checkpoints (sliding window), so 1000 can be reached
(10K limit is [set](https://github.com/apache/flink/pull/16447/files#diff-8680a6a315910794f0486c3d4234090317573c5335288b28b812f7d03c6e1b84R30) when creating the histogram).
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: issues-unsubscribe@flink.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org