This seems to be at cross-odds with most use cases, I feel. If a directory or file that I have specifically configured Logstreamer to watch simply doesn't exist, I'd rather an error be thrown than Heka sit there waiting (possibly forever) for the file to become available.
-- Robison Jacka / OpsEng / Weebly On Thu, Jan 7, 2016 at 12:35 AM, Eric LEMOINE <[email protected]> wrote: > > Le 6 janv. 2016 20:02, "Tony Cesaro" <[email protected]> a écrit : > > > > This sounds like something that could easily be solved with a > configuration management solution that builds Heka's configuration from > templates based on logic defined within it. You could check whether the > host being deployed to has the directory you are wanting to check for the > existence of and then define conditional statements to include or exclude > the segments in Heka's configuration that define Logstreamer inputs for > them. Since heka can include all files within a given directory, you could > also optionally include the entire file that defines the Logstreamer > configuration. I have had much success with something similar using > Ansible and the Jinja2 templating that it utilizes. This would, of course, > avoid changes to the Heka code altogether. > > Yes, I know this can be addressed with a configuration management > solution. Our use-case is the following: run Heka in Docker containers from > a pre-configured Heka Docker image. The same Docker image should run on > different nodes (with different apps and workloads). I don't think this is > an exotic use-case, and I think that the issue will come up again to other > people. I believe my patch provides a simple solution the that issue, hence > my insistence on having the Heka developers consider it :) > > Thanks, > > _______________________________________________ > Heka mailing list > [email protected] > https://mail.mozilla.org/listinfo/heka > >
_______________________________________________ Heka mailing list [email protected] https://mail.mozilla.org/listinfo/heka
