Hi David, no worries and thanks for the feedback!
While I agree that the "scope" feels more intuitive, and I think in the first version that's the name I used. However the issue is that there is already a pre-existing naming convention in Flink, where "variable" [1] and "additional variables" [2] names are already used for the exact thing that I'm adding here. While "scope" [3] is used for a higher level concept. Scope is for example "Task" in the "Task.numRecordsIn" metric or "TaskManager.Status.JVM.CPU" in the "TaskManager.Status.JVM.CPU.Load" metric. Metric variables are those extra tags attached to the reported metric. Like `<job_id>`, `<task_name>`. "additional variables" that user can configure in the config for each metric/trace reporter, is commonly used to add variables for things like cloud region (us_west, central_europe etc) or some other tags that are constant for the whole Flink Cluster. > If the main use case for this is to put the table name as the metric scope – could we not do this > automatically in the code without any additional need to add code? Maybe that could be some valuable follow up work. Especially since I'm not very familiar with those areas of the code myself. And AFAIK it's very common for users to use custom table registries/meta stores. Anyway there are also other good potential use cases, also in the DataStream API, so either way I think it would be good to start with this FLIP Best, Piotrek [1] https://nightlies.apache.org/flink/flink-docs-master/docs/ops/metrics/#system-scope [2] https://nightlies.apache.org/flink/flink-docs-master/docs/deployment/metric_reporters/#scope-variables-additional [3] https://nightlies.apache.org/flink/flink-docs-master/docs/ops/metrics/#scope pt., 6 gru 2024 o 12:34 David Radley <david_rad...@uk.ibm.com> napisał(a): > Hi, > Sorry I am late to this. I have some comments. > > If I am understanding this correctly, the custom metric is not a new > metric that has been customized, it is the ability to add readable useful > names / tags as scopes to existing metrics. If this is the case I suggest > changing the method name from “ .addMetricVariable(” to something like > “.addMetricScope(“ and docs should refer to metric scope / tag. The name > metric variable, to me, implies that it is a variable that is measuring > something. > > I assume these variables would be most useful if they were unique; is > there a way to make enforce uniqueness. > > From your example, I assume all the “table_name” could be any literal. If > the main use case for this is to put the table name as the metric scope – > could we not do this automatically in the code without any additional need > to add code? > > Kind regards, David. > > > From: Yuepeng Pan <panyuep...@apache.org> > Date: Friday, 6 December 2024 at 10:27 > To: dev@flink.apache.org <dev@flink.apache.org> > Subject: [EXTERNAL] Re: [VOTE] FLIP-484: Add custom metric variables to > operators > +1 (non-binding) > > Thanks Piotr for driving it ! > > Best, > Yuepeng Pan > > > On 2024/12/06 09:19:46 Piotr Nowojski wrote: > > Hi all! > > > > I would like to open the vote for FLIP-484 [1]. It has been discussed > here > > [2]. > > > > The vote will remain open for at least 72 hours (excluding the weekend). > > > > Best, > > Piotrek > > > > [1] https://cwiki.apache.org/confluence/x/44yMEw > > [2] https://lists.apache.org/thread/1g1n2l8bc27zvw88ljw4l2gqzx6rg5zy > > > > Unless otherwise stated above: > > IBM United Kingdom Limited > Registered in England and Wales with number 741598 > Registered office: Building C, IBM Hursley Office, Hursley Park Road, > Winchester, Hampshire SO21 2JN >
