U++ forum: Welcome to the forum

Re: Skylark timer jobs... [message #40924 is a reply to message #40922]

Tue, 08 October 2013 19:55

mirek
Messages: 13980
Registered: November 2005

Ultimate Member

dolik.rce wrote on Tue, 08 October 2013 12:13

mirek wrote on Mon, 07 October 2013 09:23

Well, do we really need to communicate commands back?

I mean, jobs started before prefork can stay in the main process. Each forked process then can start its own timer queue. The only thing impossible then would be killing job created in the main process in forked process or job id that is somehow transfered (via IPC) from another process. But there is a simple solution to that: make it illegal Smile

Wouldn't separate timer in each preforked process could cause same job running in each process?

Well, of course, you would have to somehow make sure that jobs scheduled before prefork are not inherited to child processes. Other than that, I think everything is OK (or not? Smile

Quote:

What do you think about the solution I proposed in RM? Do you have any specific arguments against this?

Well, perhaps, but it sounds quite complicated and 'feels' error-prone.

Quote:

mirek wrote on Mon, 07 October 2013 09:23

Related question (issue?): for the task you suggest (deleting obsolete sessions), I would normally use standard cron job. I mean I really do not see advantage of proposed job queue to normal cron job. Is there any?

Simplicity Smile

Many users (including me) like that U++ allows you to deploy whole application by deploying single executable.

Thats not mutually exclusive. We run many cron jobs in our server environment. All are implemented with single binary, which has many modes represented on commandline... That way, all code is easily shared.

Mirek

Report message to a moderator