Re: Testing with openMosix 2.6.17
"Carl Radford (DC0854-IT)" <[email protected]> Mon, 30 Oct 2006 04:12:38 -0500
| Newsgroups | gmane.linux.cluster.openmosix.general |
|---|---|
| Message-ID | <34213FE2994EDB118795000BCD6B1EB11CF3AE@corp01uk.bordersgroupinc.com> |
Hi Martin, I think you may be misconstruing what the userspace daemon does (or doesn't). It basically has nothing to do with migration other that to choose which process to migrate, and which node to migrate it to. All migration itself (the clever stuff) is handled by the kernel. It is pointless running the daemon until you have manual migration working as all it will achieve is to confusticate what is really going on (if the kernel gets it wrong, the daemon "really" gets it wrong, making a worse mess, e.g.. trying to migrate a single process to all of the cluster nodes at the same time). I have written in a dirty fix for the problems with the 2.6.17 kernel reporting a process as being "home" when it is actually migrating. If you would like it, let me know when manual migration is working and I'll stop hacking the source and let you have the latest....... -----Original Message----- From: Deadpan110 [mailto:[email protected]] Sent: 28 October 2006 22:34 To: Florian Delizy; Dirk Eddelbuettel; Carl Radford (DC0854-IT) Cc: openmosix-general-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f@public.gmane.org Subject: Re: [openMosix-general] Testing with openMosix 2.6.17 Ohh yeah... sorry about posting the patch within the mailing lists... I should have really known better... I guess my enthusiasm got the better of me... Anyways... I have been using 2.6.17-om26-20061019 for over a week now and upon re-examining the patch I created (without knowing much in the way of its workings) I noticed that there is a lot of stuff included in there that isnt oM related... was the kernel patched with updates or other bits n pieces that caused the patch to get a lil bloaty? I have noticed that using Carl Radford's migration daemon does not work on this version of oM but worked on the 2.6.17 sources from snark (git) and I think I remembered them from 2.6.15 (way back in the mailing list). I have now tried manual migration using the usual awk 'BEGIN {for(i=0;i<10000;i++)for(j=0;j<10000;j++);}' & then after looking in ps aux echo [nodeIP] >> /proc/[NUM]/om/where and it seems to hang with kcomd at 100% cpu on both nodes. The kernels are the same across nodes and compiled with CONFIG_OPENMOSIX=y CONFIG_OPENMOSIX_VERBOSE=y CONFIG_OPENMOSIX_MIGRATION_VERBOSE=y CONFIG_OPENMOSIX_DEBUG=y CONFIG_OPENMOSIX_MIGRATION_DEBUG=y # CONFIG_OPENMOSIX_DEBUG_FS is not set CONFIG_OPENMOSIX_CTRL_FS=y After having to reboot to regain access to my nodes, I am ready once again to break them :) Kernel is currently compiled for pentium pro SMP - if anyone requires me to break it some more, or recompile another... or even require me to obtain some more useful debugging info (with some basic instructions) etc... Im still available!! Martin On 24/10/06, Florian Delizy <[email protected]> wrote: > > > I have sucessfully compiled linux-2.6.17-om26-20061019 from > > http://openmosix.unfreeze.net/releases and it seems to be running > > stable although I have not yet tested manual migration. > Well, at this point I guess only the manual migration should work > BTW, I reorganized this place and I'll upload some tools as well (I > should make another release in the week fixing many issues...) > > > An initial test of Carl Radford's usertools do not appear to work with > > this version, I havent had time to do much else as Im still messing > > with setting up all my nodes - maybe recompiling the tools will give > > better results. > Well, for what version are those tools made ?? 2.4 and 2.6 are not > compatible from the user tool point of view, I would need to have a look > at the code to say exactly what to change, but you can have a look at > > http://openmosix.snarc.org > > for some description of the oM2.6 functioning > > It appears that doing: > > awk 'BEGIN {for(i=0;i<10000;i++)for(j=0;j<10000;j++);}' & > > several times, wants to migrate... but doesn't... > How is the migration order passed ? (this process should migrate)... > > I wll try it again later and get a dump of the output. > > > I have also knocked together a patch of the above oM kernel for use > > with the plain linux-2.6.17 sources. > Thanks, although I could have made one :) (I'll do for next release, I > swear), can you sent it back to me, I'll put it online. > > > > NOTE TO USERS, as you are proberly aware... oM 2.6 is still in > > development, oM 2.4.26 is still very much stable and very complete - > > with userspace tools... > > > Use the included patch for 2.6.17 AT YOUR OWN RISK! > Don't worry, we'll be stable (beta?) soon (I hope at least :) ) > > I should perhaps join the oM development lists and post there. > Well, some traffic on the general list is good as well, I'll remember to > post here anytime to announce another alpha test release :) (yes it is > alpha for now :) ) > > > Florian Delizy > > > ------------------------------------------------------------------------- > Using Tomcat but need to do more? Need to support web services, security? > Get stuff done quickly with pre-integrated technology to make your job easier > Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo > http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642 > _______________________________________________ > openMosix-general mailing list > openMosix-general-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f@public.gmane.org > https://lists.sourceforge.net/lists/listinfo/openmosix-general > ------------------------------------------------------------------------- Using Tomcat but need to do more? Need to support web services, security? Get stuff done quickly with pre-integrated technology to make your job easier Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642