Re: Judge downtime protection ideas
Ken Kleinman <[email protected]>
| Newsgroups | gmane.games.diplomacy.judge.dp.discussion |
|---|---|
| Message-ID | <[email protected]> |
Good luck with this, Manus. It sounds tricky. Comments below. On Sat, Jun 27, 2009 at 2:42 PM, Manus Hand<[email protected]> wrote: > > > DPjudgesters -- > > As you guys know (in some cases painfully), the DPjudge doesn't currently > detect the end of an outage, which means that unintentional CD's (another > reason I hate the rule) and player dismissals often happen when the judge > comes back up after it's taken a hit. > > Well, I was thinking that for the still-in-test next version of the DPjudge, > once floc.net gets back this time, I might try to put in some downtime > protection. Here's the thoughts, which I'd like comment on: > > 1. When the job that processes games runs, it will detect when the LAST > time it ran was, and if that time is more than X ago (suggestions for X? > Two hours? Twelve hours? How long constitutes an "outage"?), it will > consider an outage to have occurred. I suggest that you test for how many deadlines have passed since the last processing, and it it's greater than 1 (for any game) then assume an outage has occurred. Globally, not by game. Alternative, harder: set up a process on another machine to ping the judge. If it fails (3 times in 15 minutes?) then the judge is down, and the process sets the flag that institutes the changes resulting from item 2 and 3 discussion, below. All games check the flag before they process a turn, so turns are never processed if a deadline passed during an outage. > > 2. For those games being processed in the run just after "an outage has > occurred", if the game would process under CD, or would dismiss any players > for missing the deadline, rather than doing so, the deadline will be > extended automatically to the current date-time plus Y (suggestions for Y? > One "movement phase deadline" length?) I suggest that you take the length of time between the first deadline that occurred after the outage started, and double it, starting from the up time. > > 3. All other games would process as normal. Question: is this > insufficient? Should ALL games get their deadlines extended so that a game > that might happen to have its deadline only an hour after the judge is > restored to service don't get the shaft during an upcoming processing-run? > Perhaps the treatment should be given to all games having a deadline during > the upcoming 24 hours or anytime before?? My $.02: all games should be delayed. Ken > > Comments? > > Manus > ------------------------------------