On Sun, 2 Dec 2012 14:46:34 +1100 Dave <[email protected]> said: > In the year 2012, of the month of December, on the 1st day, Cedric BAIL > wrote: > > On Sat, Dec 1, 2012 at 12:20 PM, rustyBSD <[email protected]> wrote: > > > I took 10 minutes to create a ticket, after 4 unsuccessful > > > attemps, because > > > > > > SPANK SPANK SERVER IS DOWN > > > > Yes, e2 is almost completely overloaded. We have a new beefy server, > > e5, waiting to be configured and get service migrated to it, but > > nobody seems to have the time to do it. > > I'm an experienced unix and database admin. I could volunteer to move the > service over, if you like. That is, if you trust a maillist lurker like me. > I trust myself, if that's any help. ;)
actually we could do with people to help with: 1. server maintenance 2. website maintenance and improvement. we really need a "web team". right now we are setting up what we call "e5". e5 (e5.enlightenment.org) is our new server. it's sitting in osuosl. beber just re-installed it with gentoo and set up some stuff. the goal is to use kvm+qemu for vm's. we have a dedicated ip addr for the vms vs the host (e5v.enlightenment.org for the vm's to share). we can argue what vm to use all day - but the decision for kvm+qemu is based on the fact that osuosl ALSO offer a vm host farm that they offered us to use if we want, so we want our vm's to already be compatible with it if we choose to make use of it. sure - it may not be the "most efficient" but it'll be good enough... the hardware is good. the new server is: 2x 2.2ghz E5's (so 8 real cores, 8 threads across 2 full cpu's) 48gb ram 2x60gb ssd's raid1'd (full redundancy, 2x read rates) 4x1tb sata drivers (should be raid10'd into 1x2gb lvm, used to be, not currently - beber needs to re-format/set it up - so we should get full redundancy and 2x read rates) 1xGbit eth this is the host on which we will run vm's. the host will do basically nothing but provide data storage, memory and compute power. once its set up/working we'll not touch it and just keep it security updated where needed. no actual services will run on it.. they will ALL run on vm's now qemu has a "host fs mount" (-virtfs i think) that allows u to expose the FS from the host o the vm client. this should provide for nice ability to share files directly between vm's and a way to easily "Expand data for a vm"... so a vm ends up a vm rootfs image + launch config (ip etc.) + maybe some host fs dir archive that it may also share with others. why vm's? after years we've ended up screwing with our server and it ending up a bit of a mess. never again. we can make vm's and throw them away when done. we can install any distro we like. get tired of it? make a new vm with a different distro and bring it online, then just re-jig routing and presto. we can also now perform regular suspend/resumes and snapshots for backups. we can take those backups and copy them to LOCAL desktop machines at home and even run them. we can create and set up these vm's "at home" and upload when ready. this will lay the basics for adding vm's for bsd, windows, osx and other os's we can "get to run in kvm/qemu". we can even run ARM vm's (of course totally emulated and slow - but hey... we can have them). this allows for us to make bigger build and test farms and when/if needed we can punt off vm's into osuosl's vm cloud. so what do we need? well first we need a router vm. all network has to be routed here as we need to share port 80 and thus have to reverse-squid (haproxy) it to N possible vm's hidden behind a "inside host machine nat". this is the onyl way we can share port 80 sanely. otherwise we also need to set up ip/port forwarding for other ports and services to specific vm's. this router vm needs to go up. then we need to make a vm with the current www site - just www. not trac etc. we then need a vm with CURRENT trac+svn set up and working so we can bring it online and keep working. we'd be smart to make a SEPARATE download vm just to make it easier to isolate/track. it'd be nice to mode www.e.org/ss to a vm for user provided content (along with exchange etc.). we need a new vm with git + phabricator for moving to git. when this is ready, bring it online and then look at migrating projects from svn+trac to git+phab. when we have migrated it all over, we decommission trac+svn vm and never see it again. we need a buildbot vm. we need a doc generating/hosting vm (building docs requires we compile all of efl fyi and so it needs to compile efl ANd then install the compiled docs too). we can set up vm's for other os builds etc. as needed... we could balance out some services too. e1 and e2 (our 2 very old servers in osuosl) will go away, but e3 & e4 are 2 more we have running in france. they currently do a bunch of buildbot work. we can have them do more. they are just very limited in storage space. nb - exactly what vm's come up and what they look like is a flexible detail, but we NEED a router vm. we NEED a trac+svn vm, we NEED a git+phab vm, then we need a docs vm. everything else is pretty much fluid. it could live on the docs vm. i don't much care. so... that's here as a "blurt out the "understanding of plans". we do intend to move to git+phab - we just need a transition. we need to move over to our new server first though. the old one (e2) is likely to die at any time. it's been going without anything like a disk crash for like over 6 years... :) its raid controller is having problems (battery backup is unhappy). :) so if you can help - please talk with beber (on the list here for example) and make plans. if the above sounds vaguely sensible - then find. if its not - debate, though there has been a lot. i dont want to make things more complex than needed. we dont need to over-engineer things. we just need to be in a situation where we no longer screw up the one and only server install we have (vm's as explained above) and can easily decommission old ones and bring up new ones and migrate as needed. > > > Now that it works I wrongly hoped that I could join the backtrace, > > > and now it tells me > > > > > > Submission rejected as potential spam > > > > > > So, here is the backtrace. > > > > Ah ah ah. He has some spirit, mike did you do that to get less bug > > report ? :-) > > -- > > Cedric BAIL > > > > ------------------------------------------------------------------------------ > > Keep yourself connected to Go Parallel: > > INSIGHTS What's next for parallel hardware, programming and related areas? > > Interviews and blogs by thought leaders keep you ahead of the curve. > > http://goparallel.sourceforge.net > > _______________________________________________ > > enlightenment-devel mailing list > > [email protected] > > https://lists.sourceforge.net/lists/listinfo/enlightenment-devel > > ------------------------------------------------------------------------------ > Keep yourself connected to Go Parallel: > DESIGN Expert tips on starting your parallel project right. > http://goparallel.sourceforge.net/ > _______________________________________________ > enlightenment-devel mailing list > [email protected] > https://lists.sourceforge.net/lists/listinfo/enlightenment-devel > -- ------------- Codito, ergo sum - "I code, therefore I am" -------------- The Rasterman (Carsten Haitzler) [email protected] ------------------------------------------------------------------------------ Master HTML5, CSS3, ASP.NET, MVC, AJAX, Knockout.js, Web API and much more. Get web development skills now with LearnDevNow - 350+ hours of step-by-step video tutorials by Microsoft MVPs and experts. SALE $99.99 this month only -- learn more at: http://p.sf.net/sfu/learnmore_122812 _______________________________________________ enlightenment-devel mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/enlightenment-devel
