We are discussing the addition of exporters in #32, but it seems redundant (a watcher is very similar to an exporter) and like it might be adding a lot of bloat to the code base. But instead, couldn't we achieve the same thing by having a pushgateway watcher? For example, to create the watcher, the user might specify another watcher / one or more tasks to push to the gateway:
The above would say "use the pushgateway watcher, task function push_all_func to find any result files in watcher "results-watcher" and task "task-hpc-job" and push them to the gateway. The customization of watcher / tasks / data files would allow the user to select what data to push, and in what format.
Then for using, it would be a matter of running the tasks in serial:
We are discussing the addition of exporters in #32, but it seems redundant (a watcher is very similar to an exporter) and like it might be adding a lot of bloat to the code base. But instead, couldn't we achieve the same thing by having a pushgateway watcher? For example, to create the watcher, the user might specify another watcher / one or more tasks to push to the gateway:
The above would say "use the pushgateway watcher, task function
push_all_functo find any result files in watcher "results-watcher" and task "task-hpc-job" and push them to the gateway. The customization of watcher / tasks / data files would allow the user to select what data to push, and in what format.Then for using, it would be a matter of running the tasks in serial:
Or something along those lines :)