On Sat, Feb 18, 2017 at 2:25 PM, Glyph Lefkowitz <glyph@twistedmatrix.com> wrote:
On Feb 2, 2017, at 1:17 AM, Роман Мещеряков <romanmescheryakov@yandex.ru> wrote:
*Hello everyone! I wrote Twisted-based TCP server which is capable of running in several relatively different modes. When mode switch is needed, I would like the server to restart itself by some means, for it to read new mode from configuration file and create corresponding implementation. Also I wanted the restart to occur without re-running twistd process, because it seems to me more appropriate for Linux daemon.*
Hi Roman,
Sorry it took a while to get to this one :).
If your solution is working for you, then it's fine. You've correctly understood the API associated with IServiceCollection and your implementation (at least at first glance, I don't have time to test it exhaustively :)) is correct.
However, there's a more philosophical question as well: if you really want to *re-start* from scratch, what exactly are you shutting down and starting up again? Wouldn't you want to be able to load new code? Simply stopping and starting objects won't re-initialize the process from the beginning, only from some indeterminate middle state where parts of the program are already set up; this doesn't have the usual desired effect of a "restart" where potentially unknown or buggy state is cleared away.
You may want to consider simply calling execv https://docs.python.org/ 2.7/library/os.html#os.execv with [sys.executable] + sys.argv rather than trying to do this in terms of Twisted's APIs.
Or better yet. Don't be afraid to just exit. Your process is being managed by a supervisor with a rich feature set - like easily controlled re-restart behavior, right (If not, what do you do when your daemon simply *crashes*)? Jean-Paul