On Thu, 8 Jun 2017 22:03:28 -0700 Cedric BAIL <cedric.b...@free.fr> said:
> On Thu, Jun 8, 2017 at 7:24 PM, Andrew Williams <a...@andywilliams.me> wrote: > > "Simply not possible" is complete nonsense. We have a lot of intelligent > > people on this list and replication is not an intangible problem any more. > > How it's set up at the moment may make it harder but I think we're agreed > > that something needs to change. The concept of any organisation running all > > their infrastructure on a single host is quite laughable these days I'm > > afraid. Is it time or money that's the problem? It's certainly not > > technical limitations. > > No, it is not indeed. Mysql and gitolite do support mirroring and with > a high availability setup in front it would be perfectly doable. At > this point it is time more than money that is needed. Migrating things > to container and making them ready for high availability is highly "simply not possible" as i said. to do it you need to HA across local nodes (things with very low round trip time) because now everything you do has to become a syncrhonous transaction across 2 or more hosts. if one host is e.org and another is in some other datacenter we'll be paying for it in performance all day every day in every commit or phab interaction etc. if you don't do this and run async then you have conflict resolution issues to solve. and then the extra hosting has to be set up, configured, managed and paid for... > time consuming. The additional hosting wouldn't even be necessary as > we already have additional machine in France that can act as hot > standby. We have the hardware, internet connexion and internet > location to make this work. but then you need to duplicate the entire setup to be able to RUN from there and MAINTAIN it in a RUNNABLE state. you know how to do that? go for it... go set it up now... hurry up. it'll save our servers! everything will be fine then! :) it won't be. that's my point. we have a kernel issue because of imho a custom gentoo kernel and going off to something "well tested and used widely" should solve that. solve that and all of the anger goes away. we already have disk-level redundancy. we don't need to run full hot failover like above and pay the price across 2 machines in different locations. we really just need regular backups (rsync every hour?) and in the very very very very rare case our entire server goes down and the disks are totally inaccessible we have the "last backup" somewhere else to start up from. > Cedric > > ------------------------------------------------------------------------------ > Check out the vibrant tech community on one of the world's most > engaging tech sites, Slashdot.org! http://sdm.link/slashdot > _______________________________________________ > enlightenment-devel mailing list > enlightenment-devel@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/enlightenment-devel > -- ------------- Codito, ergo sum - "I code, therefore I am" -------------- The Rasterman (Carsten Haitzler) ras...@rasterman.com ------------------------------------------------------------------------------ Check out the vibrant tech community on one of the world's most engaging tech sites, Slashdot.org! http://sdm.link/slashdot _______________________________________________ enlightenment-devel mailing list enlightenment-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/enlightenment-devel