Did someone mention my name? :P There's not much to it, actually. What we do is:
a) instead of adjusting the hardware requirements as much as possible, we "oversize" the CA's power so that they can make a decent worker machine (in our case, it was the difference between i3 CPUs versus i7 CPUs). b) We've got a MH worker instance AND a Galicaster instance running in the machine at the same time. c) Using some sort of cron script, we use the "maintenance" endpoint in the worker to activate its maintenance mode during the day, and de-activate it at night. The times are chosen so that, should any processing work be in execution by the time the "maintenance" is triggered, there's enough time for it to complete before any recording should start (in our case, classes start at 9:00AM, so 7:00AM is an acceptable time). It's not rocket science, as you can see :). Hope it helps! Best, Rubén Pérez TELTEK Video Research www.teltek.es 2012/12/10 費納德費納德 <[email protected]> > Thank you very much for your information Greg, it is really useful. Maybe > there is an interesting solution trying to mount a worker on the capture > agent machine (or vice versa). > > I hope Ruben or someone else on Galicaster team can comment something > about that option. > > Anyway, I will start working on it and I will try to keep the community > inform about our progress. > > Best regards, > > Fernando Hernández > > > 2012/12/8 Greg Logan <[email protected]> > >> On 12-12-07 03:53 AM, 費納德費納德 wrote: >> > Hello everybody, >> > >> > we are thinking to test this solution, a machine that integrates both MH >> > elements (traditional Linux based Agent and Server). Does anybody have >> > experience on this regard? is it possible? any cons? >> >> The core is typically very CPU and disk intensive, so capturing at the >> same time that the core is processing something could be very very bad >> for both of them. Also, if your engage server is also on the same >> machine you now have *three* distinct sets of users accessing the same >> hardware in a fairly intense way. It's possible, take >> testallinone.usask.ca:8080 for example, but it requires some exceedingly >> beefy hardware. The example above also does not have a real capture >> agent, it's only using a demo capture which is just copied from source >> file to output file. >> >> I know Teltek is doing something similar with Galicaster, in that the >> CAs are used as worker nodes when they aren't capturing. You'd have to >> speak with Rubencino or someone else on their team for more information >> though, I'm not terribly familiar with their product. >> >> > My supervisor also wants we try another solution, Windows machine with >> > TheRec and a Virtual Machine with Linux OS and the Core Server, of >> > course, in the same machine. Again, the same questions, any experience? >> > cons? possible? >> >> Possible. Very difficult. We thought about something similar here at >> USask, making the host machine the CA, and running the presenter's >> machine in a VM. This was mainly scuttled by our local in-room >> technical folks who, understandably, are leery of letting us experiment >> with production equipment :) >> >> G >> >> > Best regards, >> > >> > Fernando. >> > >> > >> > _______________________________________________ >> > Matterhorn-users mailing list >> > [email protected] >> > http://lists.opencastproject.org/mailman/listinfo/matterhorn-users >> > >> >> >> >> _______________________________________________ >> Matterhorn-users mailing list >> [email protected] >> http://lists.opencastproject.org/mailman/listinfo/matterhorn-users >> >> > > _______________________________________________ > Matterhorn-users mailing list > [email protected] > http://lists.opencastproject.org/mailman/listinfo/matterhorn-users > >
_______________________________________________ Matterhorn-users mailing list [email protected] http://lists.opencastproject.org/mailman/listinfo/matterhorn-users
