You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@airflow.apache.org by GitBox <gi...@apache.org> on 2020/10/12 10:09:28 UTC

[GitHub] [airflow] potiuk opened a new issue #11463: Tags rather than names in variable parts of the metrics

potiuk opened a new issue #11463:
URL: https://github.com/apache/airflow/issues/11463


   **Description**
   
   
   It would be great if the metrics we publish in Airflow change to include TAGs for "variable" parts of those (such as dag_id, task_id, pool name) rather than being part of the metrics.
   
   See here: https://airflow.apache.org/docs/1.10.12/metrics.html, https://airflow.readthedocs.io/en/latest/logging-monitoring/metrics.html  - those variable parts should be tags.
   
   We might consider to change those metrics by default in 2.0 and possibly introduce a backport in 1.10.13 to allow people to migrate.
   
   
   **Use case / motivation**
   
   Having the variable parts of the metrics makes it really difficult to aggregate the metrics. Seems that statsd python library does not support tags, but there are possible solutions by installing extensions:
   
   https://stackoverflow.com/questions/49852654/sending-statsd-metrics-with-tags-from-python-client
   
   **Related Issues**
   
   <!-- Is there currently another issue associated with this? -->
   


----------------------------------------------------------------
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.

For queries about this service, please contact Infrastructure at:
users@infra.apache.org



[GitHub] [airflow] ashb commented on issue #11463: Tags rather than names in variable parts of the metrics

Posted by GitBox <gi...@apache.org>.
ashb commented on issue #11463:
URL: https://github.com/apache/airflow/issues/11463#issuecomment-710029386


   https://statsd.readthedocs.io/en/latest/tags.html?highlight=tag
   
   > Tagged metrics—such as those used by Datadog and Telegraf—are explicitly outside the scope of this library. Alternatives exist and are recommended. This document lays out the reasons to avoid support for tags.
   
   Yeah, so tagged metrics are not in main pystatsd client, not in "stock" statsd server.
   
   Supporting them would be good, but we possibly need to be slightly smarter about how we do it (such as convert tags in to names for fallback? Not sure)


----------------------------------------------------------------
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.

For queries about this service, please contact Infrastructure at:
users@infra.apache.org



[GitHub] [airflow] kaxil commented on issue #11463: Tags rather than names in variable parts of the metrics

Posted by GitBox <gi...@apache.org>.
kaxil commented on issue #11463:
URL: https://github.com/apache/airflow/issues/11463#issuecomment-710032223


   It would be better to natively support Prometheus (https://github.com/prometheus/client_python)


----------------------------------------------------------------
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.

For queries about this service, please contact Infrastructure at:
users@infra.apache.org



[GitHub] [airflow] ashb edited a comment on issue #11463: Tags rather than names in variable parts of the metrics

Posted by GitBox <gi...@apache.org>.
ashb edited a comment on issue #11463:
URL: https://github.com/apache/airflow/issues/11463#issuecomment-709449171


   Plain statsd doesn't support "tags" is the main issue here.
   
   At least I _thought_ it didn't. We should also see of the statsd-to-promethues supports this


----------------------------------------------------------------
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.

For queries about this service, please contact Infrastructure at:
users@infra.apache.org



[GitHub] [airflow] ashb commented on issue #11463: Tags rather than names in variable parts of the metrics

Posted by GitBox <gi...@apache.org>.
ashb commented on issue #11463:
URL: https://github.com/apache/airflow/issues/11463#issuecomment-709449171


   Plain statsd doesn't support "tags" is the main issue here


----------------------------------------------------------------
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.

For queries about this service, please contact Infrastructure at:
users@infra.apache.org



[GitHub] [airflow] potiuk commented on issue #11463: Tags rather than names in variable parts of the metrics

Posted by GitBox <gi...@apache.org>.
potiuk commented on issue #11463:
URL: https://github.com/apache/airflow/issues/11463#issuecomment-707025309


   @ashb  @kaxil  -> you have an experience with running Airflow in Production environment with Graphana dashboards etc. Do you think this one is a valid and important one ? Do you think we should target it for 2.0 and possibly backport to 1.10 for easier migration? This seems like an easy task, to not remove existing metrics but add new ones with tags instead.


----------------------------------------------------------------
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.

For queries about this service, please contact Infrastructure at:
users@infra.apache.org



[GitHub] [airflow] ashb commented on issue #11463: Tags rather than names in variable parts of the metrics

Posted by GitBox <gi...@apache.org>.
ashb commented on issue #11463:
URL: https://github.com/apache/airflow/issues/11463#issuecomment-710079872


   Native prom support would be good - but that needs a) a webserver, and b) some way of collecting metrics from across the different processes, and for components without webservers (scheduler and workers right now)


----------------------------------------------------------------
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.

For queries about this service, please contact Infrastructure at:
users@infra.apache.org