On Thu, Apr 11, 2019 at 03:12:13PM -0400, Patrick Hemmer wrote: > In regards to deprecating `stats bind-process`, I think this would be > acceptable. I can't think of any issues that might arise from that. Though > I'm not sure what else is part of this frontend, which I'm gathering is some > sort of hidden frontend used for the stats socket, master-worker > communication, and may be other things. > Seems like `stats socket ... process` would indeed be safer in that regard.
It's definitely safer, but reviewing the doc and the code I figured we also have "stats timeout" which is needed to set the frontend's timeout and "stats maxconn" to set the frontend's maxconn. This frontend is used exclusively for the CLI. But when running in master-worker mode, there is an extra socket between the master and this frontend. And it is affected by these settings, which I fear will sooner or later cause some confusion or problems. That's why in the end I suspect that we should instead split these in two, one for the normal CLI accessed via the configured stats socket, and another one dedicated to inter-process communications. Over time these two might evolve with different capabilities and this should cause less trouble I guess. A quick test with only about 10 lines changed proved to work, so we should just take a closer look and do this at least for now I think. This doesn't prevent us from reporting "stats bind-process" as deprecated though ;-) Thanks, Willy
