From burke at berkeley.edu Tue Nov 3 23:26:17 2009 From: burke at berkeley.edu (Burke Bundy) Date: Tue, 3 Nov 2009 23:26:17 -0800 Subject: [Caos] Partitioning/kickstart problem In-Reply-To: <8EF18C99-F2EC-4B9F-8657-FE71B39AE6D0@berkeley.edu> References: <8EF18C99-F2EC-4B9F-8657-FE71B39AE6D0@berkeley.edu> Message-ID: Replying to my own post... Still cannot seem to get Caos installed on this server. I get the same error (below) when using a default kickstart file: # caos-install /usr/share/caos/kickstart/server.ks Using kickstart file: /usr/share/caos/kickstart/server.ks Install preparation... Clearing partition table on /dev/ sda OK Writing 64bit disk label to /dev/ sda OK Creating partition for / boot OK ERROR: Could not identify partition for /boot! Aborting installation! Installation process complete /dev/sda is an ~ 250GB RAID1 volume and /dev/sdb is an ~ 8TB RAID5 volume. Both RAID volumes are sitting on an Adaptec RAID card. I have no problem installing CentOS on this same server, though the CentOS installer does complain about "/boot" being installed on a gpt partition. I have to drop into a virtual console and use 'parted' to write a new msdos label to /dev/sda and then the installation will proceed. I've tried doing the same within the Caos Live cd and then re-attempting a default installation but get the same error: # parted /dev/sda GNU Parted 1.8.8 Using /dev/sda Welcome to GNU Parted! Type 'help' to view a list of commands. (parted) mklabel Warning: The existing disk label on /dev/sda will be destroyed and all data on this disk will be lost. Do you want to continue? Yes/No? yes New disk label type? [gpt]? msdos (parted) print Model: SMC ARRAY 0 (scsi) Disk /dev/sda: 251GB Sector size (logical/physical): 512B/512B Partition Table: msdos Number Start End Size Type File system Flags (parted) quit Information: You may need to update /etc/fstab. [root at h2195-322-321m08-001-d ~]# caos-install /usr/share/caos/ kickstart/server.ks Using kickstart file: /usr/share/caos/kickstart/server.ks Install preparation... Clearing partition table on /dev/ sda OK Writing 64bit disk label to /dev/ sda OK Creating partition for / boot OK ERROR: Could not identify partition for /boot! Aborting installation! Installation process complete Any thoughts or advice most appreciated. Thanks, Burke On Oct 29, 2009, at 2:18 PM, Burke Bundy wrote: > Hello, > > I have a slightly modified kickstart file I'm attempting to use to > setup a new server with Caos Linux. After booting up with the Live cd > and invoking "caos-install server-custom.ks" I am greeted with the > following: > > # caos-install server-custom.ks > Using kickstart file: server-custom.ks > Install preparation... > > Clearing partition table on /dev/sda > Writing 64bit disk label to /dev/sda > Creating partition for /boot > ERROR: Could not identify partition for /boot! > > Aborting installation! > > Installation process complete > # > > Below is the "server-custom.ks" file. Any help/advice much > appreciated. > > Thanks, > Burke > > ----- > > clearpart > part /boot --size=128 --fstype=ext3 --ondisk=sda > part swap --size=2048 --ondisk=sda > part / --size=10240 --fstype=xfs --ondisk=sda > part /tmp --size=5120 --fstype=xfs --ondisk=sda > part /opt --size=40960 --fstype=xfs --ondisk=sda > part /var --size=40960 --fstype=xfs --ondisk=sda > part /usr --size=10240 --fstype=xfs --ondisk=sda > part /usr/local --size=10240 --fstype=xfs --ondisk=sda > part /srv --size=20480 --fstype=xfs --ondisk=sda > part /home --size=grow --fstype=xfs --ondisk=sdb > install --media=cdrom > bootloader > firstboot --disable > network --live > keyboard --live > timezone --live > lang --live > user sysadmin --sudo > password --user=root --live > password --user=admin --random > > > %packages --install --cdrom > metapkg-services > > %packages --update > nsa-security-check > > %services > httpd > mysql > nfs > smb > vsftpd > > > > _______________________________________________ > Caos mailing list > Caos at lists.infiscale.org > http://lists.infiscale.org/mailman/listinfo/caos From stefan at mdy.univie.ac.at Mon Nov 9 07:07:45 2009 From: stefan at mdy.univie.ac.at (Stefan Boresch) Date: Mon, 9 Nov 2009 16:07:45 +0100 Subject: [Caos] still no jmicron module for install, LiveCD fails Message-ID: <20091109150745.GU5708@loop.mdy.univie.ac.at> Finally having a chance to test the live installer/CD. Unfortunately, it still doesn't load the jmicron module, so it doesn't find the CD, and so the LiveCD doesn't help ... (please see my post from beginning of September (http://lists.infiscale.org/pipermail/caos/2009-September/002175.html)) Thus, I'm back to the NSA 1.0.8 boot disk where I know how to work around this issue ... Again, this is recent supermicro serverboard for Nehalems, so I trust this issue will bite others soon! Best regards, Stefan PS: For the record, I don't need a fully automated installer etc. But LiveCD notwithstanding, I recommend keeping the 'preinstall' option to load add. modules for exactly this type of issue, including the documentation ... -- Stefan Boresch Institute for Computational Biological Chemistry University of Vienna, Waehringerstr. 17 A-1090 Vienna, Austria Phone: -43-1-427752715 Fax: -43-1-427752790 From stefan at mdy.univie.ac.at Mon Nov 9 07:19:15 2009 From: stefan at mdy.univie.ac.at (Stefan Boresch) Date: Mon, 9 Nov 2009 16:19:15 +0100 Subject: [Caos] still no jmicron module for install, LiveCD fails In-Reply-To: <20091109150745.GU5708@loop.mdy.univie.ac.at> References: <20091109150745.GU5708@loop.mdy.univie.ac.at> Message-ID: <20091109151915.GV5708@loop.mdy.univie.ac.at> Some additional info: when the LiveCD installer doesn't find the CDROM, it drops you to a minimal shell. If one then loads the jmicron module by hand, and repeats /init , then the CD rom is found; however, it still fails later on. switch_root failed: not rootfs (I am not surprised, since certain stages of the boot sequence are most likely not intended to be carried out twice ...) I think this emphasizes my point that there needs to be a controlled and documented way *very early on* in the boot sequence to interupt the main flow of the install (/init) in order to e.g. load add. modules and try to fix things by hand, after which the normal flow of /init can be resumed ... Best regards, Stefan On Mon, Nov 09, 2009 at 04:07:45PM +0100, Stefan Boresch wrote: > Finally having a chance to test the live installer/CD. Unfortunately, > it still doesn't load the jmicron module, so it doesn't find the CD, > and so the LiveCD doesn't help ... (please see my post from > beginning of September > (http://lists.infiscale.org/pipermail/caos/2009-September/002175.html)) > > Thus, I'm back to the NSA 1.0.8 boot disk where I know how to work around > this issue ... Again, this is recent supermicro serverboard for Nehalems, > so I trust this issue will bite others soon! > > Best regards, > > Stefan > > PS: For the record, I don't need a fully automated installer etc. But > LiveCD notwithstanding, I recommend keeping the 'preinstall' option to > load add. modules for exactly this type of issue, including the > documentation ... > > -- > Stefan Boresch > Institute for Computational Biological Chemistry > University of Vienna, Waehringerstr. 17 A-1090 Vienna, Austria > Phone: -43-1-427752715 Fax: -43-1-427752790 > _______________________________________________ > Caos mailing list > Caos at lists.infiscale.org > http://lists.infiscale.org/mailman/listinfo/caos > -- Stefan Boresch Institute for Computational Biological Chemistry University of Vienna, Waehringerstr. 17 A-1090 Vienna, Austria Phone: -43-1-427752715 Fax: -43-1-427752790 From gmk at infiscale.org Mon Nov 9 10:50:38 2009 From: gmk at infiscale.org (Greg Kurtzer) Date: Mon, 9 Nov 2009 10:50:38 -0800 Subject: [Caos] still no jmicron module for install, LiveCD fails In-Reply-To: <20091109151915.GV5708@loop.mdy.univie.ac.at> References: <20091109150745.GU5708@loop.mdy.univie.ac.at> <20091109151915.GV5708@loop.mdy.univie.ac.at> Message-ID: <20091109185036.GA7472@infiscale.org> Hi Stefan, The preshell boot option should exist and work fine. In a nutshell can you try to boot with the preshell option and then "modprobe jmicron; exit" and see if that works for you? I am not familiar with the jmicron module. What hardware is it supporting (is it the CDROM itself, or a bus, or PCI/bridge device)? Thanks, Greg On Monday, 09 November 2009, at 16:19:15 (+0100), Stefan Boresch wrote: > Some additional info: > > when the LiveCD installer doesn't find the CDROM, it drops you to a > minimal shell. If one then loads the jmicron module by hand, and repeats > /init , then the CD rom is found; however, it still fails later on. > > switch_root failed: not rootfs > > (I am not surprised, since certain stages of the boot sequence are most > likely not intended to be carried out twice ...) > > I think this emphasizes my point that there needs to be a controlled > and documented way *very early on* in the boot sequence to interupt > the main flow of the install (/init) in order to e.g. load > add. modules and try to fix things by hand, after which the normal > flow of /init can be resumed ... > > Best regards, > > Stefan > > On Mon, Nov 09, 2009 at 04:07:45PM +0100, Stefan Boresch wrote: > > Finally having a chance to test the live installer/CD. Unfortunately, > > it still doesn't load the jmicron module, so it doesn't find the CD, > > and so the LiveCD doesn't help ... (please see my post from > > beginning of September > > (http://lists.infiscale.org/pipermail/caos/2009-September/002175.html)) > > > > Thus, I'm back to the NSA 1.0.8 boot disk where I know how to work around > > this issue ... Again, this is recent supermicro serverboard for Nehalems, > > so I trust this issue will bite others soon! > > > > Best regards, > > > > Stefan > > > > PS: For the record, I don't need a fully automated installer etc. But > > LiveCD notwithstanding, I recommend keeping the 'preinstall' option to > > load add. modules for exactly this type of issue, including the > > documentation ... > > > > -- > > Stefan Boresch > > Institute for Computational Biological Chemistry > > University of Vienna, Waehringerstr. 17 A-1090 Vienna, Austria > > Phone: -43-1-427752715 Fax: -43-1-427752790 > > _______________________________________________ > > Caos mailing list > > Caos at lists.infiscale.org > > http://lists.infiscale.org/mailman/listinfo/caos > > > > -- > Stefan Boresch > Institute for Computational Biological Chemistry > University of Vienna, Waehringerstr. 17 A-1090 Vienna, Austria > Phone: -43-1-427752715 Fax: -43-1-427752790 > _______________________________________________ > Caos mailing list > Caos at lists.infiscale.org > http://lists.infiscale.org/mailman/listinfo/caos -- Greg M. Kurtzer Chief Technology Officer HPC Systems Architect Infiscale, Inc. - http://www.infiscale.com From stefan at mdy.univie.ac.at Mon Nov 9 12:54:37 2009 From: stefan at mdy.univie.ac.at (stefan at mdy.univie.ac.at) Date: Mon, 9 Nov 2009 21:54:37 +0100 (CET) Subject: [Caos] still no jmicron module for install, LiveCD fails In-Reply-To: <20091109185036.GA7472@infiscale.org> References: <20091109150745.GU5708@loop.mdy.univie.ac.at> <20091109151915.GV5708@loop.mdy.univie.ac.at> <20091109185036.GA7472@infiscale.org> Message-ID: <1118.80.123.53.42.1257800077.squirrel@www.mdy.univie.ac.at> Hi Greg, > The preshell boot option should exist and work fine. In a nutshell can you > try to boot with the preshell option and then "modprobe jmicron; exit" and > see if that works for you? OK, will try that tomorrow. However, if preshell works as in 1.0.8 then it won't suffice since the installer attempts to locate the CDROM before the preshell kicks in. But, I'll try. > > I am not familiar with the jmicron module. What hardware is it supporting > (is it the CDROM itself, or a bus, or PCI/bridge device)? Well, I wasn't until September, but the hardware tech support guy who helped me sort this out then told me that jmicron (and consorts) are becoming more and more frequent. Basically, if I am getting this right, modern Mobos are SATA2 (and better) only (as far as the BIOS is concerned). Most CD/DVD drives, however still are ATA. The BIOS in this Supermicro Server board is too "noble" to talk ATA, but it lets a "BIOS module" probe whether there is need for ATA support. (During boot this looks very similar to a SCSI BIOS probing the presence of drives, or a 3ware RAID controller looking for disks under its control) The normal BIOS, however, doesn't report the presence of the CD/DVD to the OS; the OS has to load the necessary support to talk to this BIOS extension ... (in my case the jmicron module) Don't know whether this makes any sense .. To me this is reminiscent of the 90s when we needed zillions of CDROM drivers ... More tomorrow, Stefan From gmk at infiscale.org Mon Nov 9 14:12:03 2009 From: gmk at infiscale.org (Greg Kurtzer) Date: Mon, 9 Nov 2009 14:12:03 -0800 Subject: [Caos] still no jmicron module for install, LiveCD fails In-Reply-To: <1118.80.123.53.42.1257800077.squirrel@www.mdy.univie.ac.at> References: <20091109150745.GU5708@loop.mdy.univie.ac.at> <20091109151915.GV5708@loop.mdy.univie.ac.at> <20091109185036.GA7472@infiscale.org> <1118.80.123.53.42.1257800077.squirrel@www.mdy.univie.ac.at> Message-ID: <20091109221203.GB7472@infiscale.org> On Monday, 09 November 2009, at 21:54:37 (+0100), stefan at mdy.univie.ac.at wrote: > Hi Greg, > > > The preshell boot option should exist and work fine. In a nutshell can you > > try to boot with the preshell option and then "modprobe jmicron; exit" and > > see if that works for you? > > OK, will try that tomorrow. However, if preshell works as in 1.0.8 then it > won't suffice since the installer attempts to locate the CDROM before the > preshell kicks in. But, I'll try. The new installer does the preshell right before the hardware probe, so it should work. > > > > I am not familiar with the jmicron module. What hardware is it supporting > > (is it the CDROM itself, or a bus, or PCI/bridge device)? > > Well, I wasn't until September, but the hardware tech support guy who > helped me sort this out then told me that jmicron (and consorts) are becoming > more and more frequent. Basically, if I am getting this right, modern > Mobos are SATA2 (and better) only (as far as the BIOS is concerned). Most > CD/DVD drives, however still are ATA. The BIOS in this Supermicro Server > board is too "noble" to talk ATA, but it lets a "BIOS module" probe > whether there is need for ATA support. (During boot this looks very > similar to a SCSI BIOS probing the presence of drives, or a 3ware RAID > controller looking for disks under its control) The normal BIOS, however, > doesn't report the presence of the CD/DVD to the OS; the OS has to load > the necessary support to talk to this BIOS extension ... (in my case the > jmicron module) Don't know whether this makes any sense .. To me this is > reminiscent of the 90s when we needed zillions of CDROM drivers ... Ahh, very interesting. Can you make a bug report requesting the addition of this module at http://bugs.infiscale.org so we don't miss it on our next round of installers? Thanks for the info! Greg -- Greg M. Kurtzer Chief Technology Officer HPC Systems Architect Infiscale, Inc. - http://www.infiscale.com From stefan at mdy.univie.ac.at Tue Nov 10 02:32:31 2009 From: stefan at mdy.univie.ac.at (Stefan Boresch) Date: Tue, 10 Nov 2009 11:32:31 +0100 Subject: [Caos] still no jmicron module for install, LiveCD fails In-Reply-To: <20091109221203.GB7472@infiscale.org> References: <20091109150745.GU5708@loop.mdy.univie.ac.at> <20091109151915.GV5708@loop.mdy.univie.ac.at> <20091109185036.GA7472@infiscale.org> <1118.80.123.53.42.1257800077.squirrel@www.mdy.univie.ac.at> <20091109221203.GB7472@infiscale.org> Message-ID: <20091110103231.GX5708@loop.mdy.univie.ac.at> Hi Greg, I filed a bug report in bugzilla. However, ... On Mon, Nov 09, 2009 at 02:12:03PM -0800, Greg Kurtzer wrote: > On Monday, 09 November 2009, at 21:54:37 (+0100), > stefan at mdy.univie.ac.at wrote: > The new installer does the preshell right before the hardware probe, so it > should work. ... I seem to be very dumb, but I find no way to let the system know that I want a 'preshell'. I get the menu as shown in the LiveCD/Installer walkthrough, but I have no possibility to pass options ... I should maybe add that I am using the -base-1.0.25 image, since I have no tool to burn 2GB to a CD (not clear what the 'full' image is all about -- seems to fit DVDs only) I mention this since in my experience for 1.0.8, the full and base images are not equivalent concerning the install process, the 1.0.8-base image/installer for me always fails to install grub, whereas the full image works fine (and, yes, I have checked md5sums!) Note that with the new installer, after the install fails, I am dropped into a shell (that's better than with the old installer, where at this point you were 'dead'), and I can load the jmicron module at this point. However, I don't know for the new installer how I could repeat the steps starting from looking for the CD-ROM ... Best regards, Stefan -- Stefan Boresch Institute for Computational Biological Chemistry University of Vienna, Waehringerstr. 17 A-1090 Vienna, Austria Phone: -43-1-427752715 Fax: -43-1-427752790 From gmk at infiscale.org Tue Nov 10 10:44:39 2009 From: gmk at infiscale.org (Greg Kurtzer) Date: Tue, 10 Nov 2009 10:44:39 -0800 Subject: [Caos] still no jmicron module for install, LiveCD fails In-Reply-To: <20091110103231.GX5708@loop.mdy.univie.ac.at> References: <20091109150745.GU5708@loop.mdy.univie.ac.at> <20091109151915.GV5708@loop.mdy.univie.ac.at> <20091109185036.GA7472@infiscale.org> <1118.80.123.53.42.1257800077.squirrel@www.mdy.univie.ac.at> <20091109221203.GB7472@infiscale.org> <20091110103231.GX5708@loop.mdy.univie.ac.at> Message-ID: <20091110184439.GC7472@infiscale.org> Hi Stefan, Sorry about not explaining that clearly. At the CDROM boot menu press [ESC] and that will give you a "boot: " prompt. From there do: boot: livemedia preshell when it boots it will give you the shell. From there, type: # modprobe jmicron # ... any other commands you want ... # exit It will then continue booting. Hope that helps! Greg On Tuesday, 10 November 2009, at 11:32:31 (+0100), Stefan Boresch wrote: > Hi Greg, > > I filed a bug report in bugzilla. However, ... > > On Mon, Nov 09, 2009 at 02:12:03PM -0800, Greg Kurtzer wrote: > > On Monday, 09 November 2009, at 21:54:37 (+0100), > > stefan at mdy.univie.ac.at wrote: > > The new installer does the preshell right before the hardware probe, so it > > should work. > > ... I seem to be very dumb, but I find no way to let the system know that > I want a 'preshell'. I get the menu as shown in the LiveCD/Installer > walkthrough, but I have no possibility to pass options ... > > I should maybe add that I am using the -base-1.0.25 image, since I > have no tool to burn 2GB to a CD (not clear what the 'full' image is > all about -- seems to fit DVDs only) I mention this since in my > experience for 1.0.8, the full and base images are not equivalent > concerning the install process, the 1.0.8-base image/installer for me > always fails to install grub, whereas the full image works fine (and, > yes, I have checked md5sums!) > > Note that with the new installer, after the install fails, I am > dropped into a shell (that's better than with the old installer, where > at this point you were 'dead'), and I can load the jmicron module at > this point. However, I don't know for the new installer how I could > repeat the steps starting from looking for the CD-ROM ... > > Best regards, > > Stefan > > -- > Stefan Boresch > Institute for Computational Biological Chemistry > University of Vienna, Waehringerstr. 17 A-1090 Vienna, Austria > Phone: -43-1-427752715 Fax: -43-1-427752790 > _______________________________________________ > Caos mailing list > Caos at lists.infiscale.org > http://lists.infiscale.org/mailman/listinfo/caos -- Greg M. Kurtzer Chief Technology Officer HPC Systems Architect Infiscale, Inc. - http://www.infiscale.com From stefan at mdy.univie.ac.at Wed Nov 11 00:57:20 2009 From: stefan at mdy.univie.ac.at (Stefan Boresch) Date: Wed, 11 Nov 2009 09:57:20 +0100 Subject: [Caos] still no jmicron module for install, LiveCD fails In-Reply-To: <20091110184439.GC7472@infiscale.org> References: <20091109150745.GU5708@loop.mdy.univie.ac.at> <20091109151915.GV5708@loop.mdy.univie.ac.at> <20091109185036.GA7472@infiscale.org> <1118.80.123.53.42.1257800077.squirrel@www.mdy.univie.ac.at> <20091109221203.GB7472@infiscale.org> <20091110103231.GX5708@loop.mdy.univie.ac.at> <20091110184439.GC7472@infiscale.org> Message-ID: <20091111085720.GZ5708@loop.mdy.univie.ac.at> OK, success (mostly) ... On Tue, Nov 10, 2009 at 10:44:39AM -0800, Greg Kurtzer wrote: > > Sorry about not explaining that clearly. At the CDROM boot menu press [ESC] > and that will give you a "boot: " prompt. From there do: > > boot: livemedia preshell > this worked with a slight snag: Once in the preshell environment (which is indeed entered before the hardware probe etc), no more (USB) keyboard. I have seen this with other machines in the provisioning stage of perceus (i.e., before the real system comes up); on these machines I could work around by enabling "legacy support for USB" in the BIOS. However, on this machine, I cannot find such an option in the BIOS (Supermicro X8DTN+, Ami BIOS v. 2.67) An ancient non-USB keyboard works just fine. At that point everything works as expected (modprobe jmicron, exit, CD now found and install continues normally). Should I file a bug report for this ? Best regards, Stefan -- Stefan Boresch Institute for Computational Biological Chemistry University of Vienna, Waehringerstr. 17 A-1090 Vienna, Austria Phone: -43-1-427752715 Fax: -43-1-427752790 From stefan at mdy.univie.ac.at Wed Nov 11 01:28:06 2009 From: stefan at mdy.univie.ac.at (Stefan Boresch) Date: Wed, 11 Nov 2009 10:28:06 +0100 Subject: [Caos] Some comments on new LiveCD installer Message-ID: <20091111092806.GA5708@loop.mdy.univie.ac.at> Since I now finally had the opportunity to install from the new live CD, I take the liberty to comment on my experiences ( I am by no means an expert, but my first install of CaosNSA was with a pre 1.0 version, so I have experienced the 'evolution' of the installer a bit) * As outlined in the "jmicron thread", I needed the 'preshell' option to load an additional module. Is this documented *anywhere*? It was documented in early versions of the old installer, but then became an undocumented feature. The old installer entered preshell "too late" (after hardware probing and looking for the install CD). The LiveCD installer enters preshell before even probing the hardware. Side effect for me, no USB keyboard (old normal keyboard worked fortunately...) To throw in my two cents: First, things like preshell need to be documented. Second, just as the old preshell came too late, the new preshell seems to come a tad (too) early. At least an USB keyboard should be usable at this point ... ;-) I guess for me the ideal point in the boot sequence would be just after hardware probing and autoloading of modules, but *before* anything else! Of course, this might cause trouble if the autoprobe chose a wrong module or one wanted to make sure that a specific module is loaded first ... Just curious: is there a way to pass the name of a module (or of modules) to be loaded from the boot: command line? * For someone intending to install caos anyways, the benefits of the new installer are meager. I would argue that the old installer with final install after reboot was at least as quick (or course, I typed answers almost blindly). And mostly, one sees the same text based interface (which is fine for me!) * There is some flexibility lost -- how can I tell the installer to enable manual partitioning or preset network configs? * I prefer the sidekick menu of 'profiles' to the new (fewer) choices offered by the LiveCD installer * Who reads documentation ;-) I completely overlooked the admin password the first time; I only even noticed the presence of the admin account when looking at the documentation because I missed the creation of an administrative account (with a name of my choice) ... Are there any special roles intended for admin (if not, on my machines it will be wiped in favor of the named admin accounts ...) * when rebooting, the CD wasn't ejected -- this is not a problem, since boot from local disk is the default, but it takes time Finally, some questions, observations: * Is there now the possibility to do net-installs (preferably completely automated with kickstart?) * My other big feature request would be support for software raid (I have some machines which I would love to have completely (including /boot) on a RAID 1 * Finally, an observation not related to the installer. Sometimes, mirror.caoslinux.org 'works' for me in the sense that I transferred to various servers, most of them in Europe, which typically are fast. At other times (right now), I download from mirror.caoslinux.org with steaming 20-30kB/s. I don't want to "hardcode" any of the European mirrors that usually are fast, because Murphy is bound to strike at some point; I am just curious whether there is an explanation for this... I hope this doesn't come over as too negative -- thanks for all the great work! Best regards, Stefan -- Stefan Boresch Institute for Computational Biological Chemistry University of Vienna, Waehringerstr. 17 A-1090 Vienna, Austria Phone: -43-1-427752715 Fax: -43-1-427752790 From gmk at infiscale.org Mon Nov 16 12:34:52 2009 From: gmk at infiscale.org (Greg Kurtzer) Date: Mon, 16 Nov 2009 12:34:52 -0800 Subject: [Caos] [Wishlist] nvidia opencl In-Reply-To: <5292fd2d0910151316n24973bc4g6017d4ae89152024@mail.gmail.com> References: <20091015065057.GA5898@loop.mdy.univie.ac.at> <20091015183954.GH7237@infiscale.org> <5292fd2d0910151316n24973bc4g6017d4ae89152024@mail.gmail.com> Message-ID: <20091116203452.GB8270@infiscale.org> Just an FYI, the Nvidia 190.42 drivers were packaged last night and are in the nsa-testing repository now (along with a new kernel that also needs some testing). Thanks, Greg On Thursday, 15 October 2009, at 13:16:00 (-0700), Ian Kaufman wrote: > It's in beta - 190.32 is the current release. > > However, since it is beta, I would not waste the cycles. > > Ian > > > > On Thu, Oct 15, 2009 at 11:39 AM, Greg Kurtzer wrote: > > > Hi Stefan, > > > > Has Nvidia released 190.x? I only see 185.x on their site. > > > > Yeah, when they do we will do an upgrade but I can't guarantee how fast > > that > > will occur. Ping me off list if you require it ASAP. > > > > Thanks, > > Greg > > > > On Thursday, 15 October 2009, at 08:50:57 (+0200), > > Stefan Boresch wrote: > > > > > I am just wondering what the chances are that the latest round > > > of nvidia drivers (190.x) will be packaged (at least in testing). > > > These are needed for OpenCL programming (as well as the latest > > > CUDA 2.3 environment) > > > > > > I understand if that's low priority; in that case I'd just install them > > > by hand ... > > > > > > Thanks, > > > > > > Stefan > > > > > > -- > > > Stefan Boresch > > > Institute for Computational Biological Chemistry > > > University of Vienna, Waehringerstr. 17 A-1090 Vienna, Austria > > > Phone: -43-1-427752715 Fax: -43-1-427752790 > > > _______________________________________________ > > > Caos mailing list > > > Caos at lists.infiscale.org > > > http://lists.infiscale.org/mailman/listinfo/caos > > > > -- > > Greg M. Kurtzer > > Chief Technology Officer > > HPC Systems Architect > > Infiscale, Inc. - http://www.infiscale.com > > _______________________________________________ > > Caos mailing list > > Caos at lists.infiscale.org > > http://lists.infiscale.org/mailman/listinfo/caos > > > > > > -- > Ian Kaufman > Research Systems Administrator > UC San Diego, Jacobs School of Engineering ikaufman AT ucsd DOT edu > _______________________________________________ > Caos mailing list > Caos at lists.infiscale.org > http://lists.infiscale.org/mailman/listinfo/caos -- Greg M. Kurtzer Chief Technology Officer HPC Systems Architect Infiscale, Inc. - http://www.infiscale.com From stefan at mdy.univie.ac.at Tue Nov 17 00:45:42 2009 From: stefan at mdy.univie.ac.at (Stefan Boresch) Date: Tue, 17 Nov 2009 09:45:42 +0100 Subject: [Caos] [Wishlist] nvidia opencl In-Reply-To: <20091116203452.GB8270@infiscale.org> References: <20091015065057.GA5898@loop.mdy.univie.ac.at> <20091015183954.GH7237@infiscale.org> <5292fd2d0910151316n24973bc4g6017d4ae89152024@mail.gmail.com> <20091116203452.GB8270@infiscale.org> Message-ID: <20091117084542.GC22698@loop.mdy.univie.ac.at> On Mon, Nov 16, 2009 at 12:34:52PM -0800, Greg Kurtzer wrote: > Just an FYI, the Nvidia 190.42 drivers were packaged last night and are in the > nsa-testing repository now (along with a new kernel that also needs some > testing). > Thank you! I noticed both yesterday and have alreayd upgraded. No problems so far, although upgrading CUDA is a pain (not your fault, god knows why nvidia changed all their paths, so if one doesn't follow their build/make hierarchy, some cursing is in order ...) There is one remaining inconsistency between mesalib-devel freeglut-devel and nvidia-devel (this was also present with the 185.x drivers, and I even believe with 180.x ...) : nvidia-devel and mesalib-devel collide (and exclude each other) since they provide identical files, e.g. /usr/include/GL/gl.h HOWEVER, only mesalib-devel provides /usr/include/GL/glu.h, which in turn (or rather mesalib-devel) is the prerequisite if one requires /usr/include/GL/glut.h (provided by freeglut-devel) Is there a way to break this circle (other than rpm -i --nodeps ...) Best regards and thanks, Stefan > Thanks, > Greg > > On Thursday, 15 October 2009, at 13:16:00 (-0700), > Ian Kaufman wrote: > > > It's in beta - 190.32 is the current release. > > > > However, since it is beta, I would not waste the cycles. > > > > Ian > > > > > > > > On Thu, Oct 15, 2009 at 11:39 AM, Greg Kurtzer wrote: > > > > > Hi Stefan, > > > > > > Has Nvidia released 190.x? I only see 185.x on their site. > > > > > > Yeah, when they do we will do an upgrade but I can't guarantee how fast > > > that > > > will occur. Ping me off list if you require it ASAP. > > > > > > Thanks, > > > Greg > > > > > > On Thursday, 15 October 2009, at 08:50:57 (+0200), > > > Stefan Boresch wrote: > > > > > > > I am just wondering what the chances are that the latest round > > > > of nvidia drivers (190.x) will be packaged (at least in testing). > > > > These are needed for OpenCL programming (as well as the latest > > > > CUDA 2.3 environment) > > > > > > > > I understand if that's low priority; in that case I'd just install them > > > > by hand ... > > > > > > > > Thanks, > > > > > > > > Stefan > > > > > > > > -- > > > > Stefan Boresch > > > > Institute for Computational Biological Chemistry > > > > University of Vienna, Waehringerstr. 17 A-1090 Vienna, Austria > > > > Phone: -43-1-427752715 Fax: -43-1-427752790 > > > > _______________________________________________ > > > > Caos mailing list > > > > Caos at lists.infiscale.org > > > > http://lists.infiscale.org/mailman/listinfo/caos > > > > > > -- > > > Greg M. Kurtzer > > > Chief Technology Officer > > > HPC Systems Architect > > > Infiscale, Inc. - http://www.infiscale.com > > > _______________________________________________ > > > Caos mailing list > > > Caos at lists.infiscale.org > > > http://lists.infiscale.org/mailman/listinfo/caos > > > > > > > > > > > -- > > Ian Kaufman > > Research Systems Administrator > > UC San Diego, Jacobs School of Engineering ikaufman AT ucsd DOT edu > > > _______________________________________________ > > Caos mailing list > > Caos at lists.infiscale.org > > http://lists.infiscale.org/mailman/listinfo/caos > > > -- > Greg M. Kurtzer > Chief Technology Officer > HPC Systems Architect > Infiscale, Inc. - http://www.infiscale.com > _______________________________________________ > Caos mailing list > Caos at lists.infiscale.org > http://lists.infiscale.org/mailman/listinfo/caos > -- Stefan Boresch Institute for Computational Biological Chemistry University of Vienna, Waehringerstr. 17 A-1090 Vienna, Austria Phone: -43-1-427752715 Fax: -43-1-427752790 From gmk at infiscale.org Tue Nov 17 19:11:42 2009 From: gmk at infiscale.org (Greg Kurtzer) Date: Tue, 17 Nov 2009 19:11:42 -0800 Subject: [Caos] [Wishlist] nvidia opencl In-Reply-To: <20091117084542.GC22698@loop.mdy.univie.ac.at> References: <20091015065057.GA5898@loop.mdy.univie.ac.at> <20091015183954.GH7237@infiscale.org> <5292fd2d0910151316n24973bc4g6017d4ae89152024@mail.gmail.com> <20091116203452.GB8270@infiscale.org> <20091117084542.GC22698@loop.mdy.univie.ac.at> Message-ID: <20091118031100.GA21856@infiscale.org> On Tuesday, 17 November 2009, at 09:45:42 (+0100), Stefan Boresch wrote: > On Mon, Nov 16, 2009 at 12:34:52PM -0800, Greg Kurtzer wrote: > > Just an FYI, the Nvidia 190.42 drivers were packaged last night and are in the > > nsa-testing repository now (along with a new kernel that also needs some > > testing). > > > > Thank you! I noticed both yesterday and have alreayd upgraded. No > problems so far, although upgrading CUDA is a pain (not your fault, > god knows why nvidia changed all their paths, so if one doesn't follow > their build/make hierarchy, some cursing is in order ...) Hehe. :P > There is one remaining inconsistency between mesalib-devel freeglut-devel > and nvidia-devel (this was also present with the 185.x drivers, and I even > believe with 180.x ...) : > > nvidia-devel and mesalib-devel collide (and exclude each other) since > they provide identical files, e.g. /usr/include/GL/gl.h HOWEVER, only > mesalib-devel provides /usr/include/GL/glu.h, which in turn (or rather > mesalib-devel) is the prerequisite if one requires > /usr/include/GL/glut.h (provided by freeglut-devel) Is there a way to > break this circle (other than rpm -i --nodeps ...) Yes, there are several conflicting files. I need to research this again to see what the best solution would be (so if anyone has any ideas, please let me know). Thanks, Greg -- Greg M. Kurtzer Chief Technology Officer HPC Systems Architect Infiscale, Inc. - http://www.infiscale.com From stefan at mdy.univie.ac.at Tue Nov 17 23:44:22 2009 From: stefan at mdy.univie.ac.at (Stefan Boresch) Date: Wed, 18 Nov 2009 08:44:22 +0100 Subject: [Caos] [Wishlist] nvidia opencl In-Reply-To: <20091118031100.GA21856@infiscale.org> References: <20091015065057.GA5898@loop.mdy.univie.ac.at> <20091015183954.GH7237@infiscale.org> <5292fd2d0910151316n24973bc4g6017d4ae89152024@mail.gmail.com> <20091116203452.GB8270@infiscale.org> <20091117084542.GC22698@loop.mdy.univie.ac.at> <20091118031100.GA21856@infiscale.org> Message-ID: <20091118074422.GI22698@loop.mdy.univie.ac.at> Greg, On Tue, Nov 17, 2009 at 07:11:42PM -0800, Greg Kurtzer wrote: > On Tuesday, 17 November 2009, at 09:45:42 (+0100), > Stefan Boresch wrote: > > > On Mon, Nov 16, 2009 at 12:34:52PM -0800, Greg Kurtzer wrote: > > nvidia-devel and mesalib-devel collide (and exclude each other) since > > they provide identical files, e.g. /usr/include/GL/gl.h HOWEVER, only > > mesalib-devel provides /usr/include/GL/glu.h, which in turn (or rather > > mesalib-devel) is the prerequisite if one requires > > /usr/include/GL/glut.h (provided by freeglut-devel) Is there a way to > > break this circle (other than rpm -i --nodeps ...) > > Yes, there are several conflicting files. I need to research this again to > see what the best solution would be (so if anyone has any ideas, please let > me know). > I am complete naive as far as packaging is concerned, and the only time I wrapped up something myself was on Debian. It seems to me that one major part of the problem is mainly that freeglut-devel requires mesalib-devel, which is the file colliding with nvidia-devel. Debian's dpkg (I believe) not necessarily looks for a specific package as prerequisite, but for something that is provided (e.g., provides gl_includes). (On a higher level, this is how they can support postfix, exim, sendmail etc. as primary mail servers, all provide "mail") If one could do something with rpm, one would be halfway there, i.e., mesalib-devel and nvidia-devel conflict, but both provide what freeglut-devel needs; hence freeglut-devel is happy once either mesalib-devel or nvidia-devel are installed (and installing nvidia-devel would silently deinstall mesalib-devel). Should of course both mesalib-devel and freeglut-devel contain files that are also in nvidia-devel, then this would not work ... My 5 cents, Stefan -- Stefan Boresch Institute for Computational Biological Chemistry University of Vienna, Waehringerstr. 17 A-1090 Vienna, Austria Phone: -43-1-427752715 Fax: -43-1-427752790 From chris.hunter at yale.edu Thu Nov 19 15:43:46 2009 From: chris.hunter at yale.edu (Chris Hunter) Date: Thu, 19 Nov 2009 18:43:46 -0500 Subject: [Caos] newbie caos questions Message-ID: <4B05D832.3020906@yale.edu> Hi, I'm just signed onto the list. Our group is looking to move from stateless to stateful provisioning of hpc clusters. We have been using perceus for a few years. I have gone over the caoslinux wiki pages for clustering, but it not clear if its stateless(ie. ramdisk) or stateful(ie. install to local disk). I know caoslinux can be installed to a local disk but can we centralize management of the local disk? Here is our ideal situation: i) new nodes are provisioned over the network (eg. pxeboot/tftp) ii) OS image/vnfs capsule is transferred over network and installed to local disk on the node as part of provisioning iii) a central "master" server/image is used for changes to a "master" OS image/vnfs capsule iv) changes to the master image are pushed out to the nodes (via rsync, xcpu, etc). We don't want to reboot a compute node to update the image/vnfs. We don't want a ramdisk for the compute node OS/vnfs install. My question is if caoslinux can handle stateful provisioning of a cluster. Regards, Chris Hunter From mej at caoslinux.org Thu Nov 19 16:29:12 2009 From: mej at caoslinux.org (Michael Jennings) Date: Thu, 19 Nov 2009 16:29:12 -0800 Subject: [Caos] newbie caos questions In-Reply-To: <4B05D832.3020906@yale.edu> References: <4B05D832.3020906@yale.edu> Message-ID: <20091120002912.GM15649@kainx.org> On Thursday, 19 November 2009, at 18:43:46 (-0500), Chris Hunter wrote: > Hi, I'm just signed onto the list. Our group is looking to move from > stateless to stateful provisioning of hpc clusters. We have been using > perceus for a few years. > > I have gone over the caoslinux wiki pages for clustering, but it not > clear if its stateless(ie. ramdisk) or stateful(ie. install to local > disk). I know caoslinux can be installed to a local disk but can we > centralize management of the local disk? > > Here is our ideal situation: > i) new nodes are provisioned over the network (eg. pxeboot/tftp) > ii) OS image/vnfs capsule is transferred over network and installed to > local disk on the node as part of provisioning > iii) a central "master" server/image is used for changes to a "master" > OS image/vnfs capsule > iv) changes to the master image are pushed out to the nodes (via rsync, > xcpu, etc). We don't want to reboot a compute node to update the > image/vnfs. We don't want a ramdisk for the compute node OS/vnfs install. > > My question is if caoslinux can handle stateful provisioning of a cluster. To quote a prominent vendor's recent keynote address: "In our experience, stateful provisioning is never the right answer." That said, Caos is a distribution, not a cluster manager. It doesn't care how it's provisioned. You can provision it statefully or statelessly. Michael -- Michael Jennings (a.k.a. KainX) http://www.kainx.org/ Linux Server/Cluster Admin, LBL.gov Author, Eterm (www.eterm.org) ----------------------------------------------------------------------- "Sometimes I give myself the creeps. Sometimes my mind plays tricks on me. It all keeps adding up; I think I'm cracking up. Am I just paranoid? Am I just stoned?" -- Green Day, "Basket Case" From astevens at infiscale.com Fri Nov 20 06:24:35 2009 From: astevens at infiscale.com (Arthur Stevens) Date: Fri, 20 Nov 2009 06:24:35 -0800 Subject: [Caos] newbie caos questions References: <4B05D832.3020906@yale.edu> <20091120002912.GM15649@kainx.org> Message-ID: Hi Chris, as a 3rd option, hybridization might be an alternative if trying to free more ram but keep the versatility (and sanity) of stateless. Caos supports them all quite well :) Arthur ----- Original Message ----- From: "Michael Jennings" To: Sent: Thursday, November 19, 2009 4:29 PM Subject: Re: [Caos] newbie caos questions > On Thursday, 19 November 2009, at 18:43:46 (-0500), > Chris Hunter wrote: > >> Hi, I'm just signed onto the list. Our group is looking to move from >> stateless to stateful provisioning of hpc clusters. We have been using >> perceus for a few years. >> >> I have gone over the caoslinux wiki pages for clustering, but it not >> clear if its stateless(ie. ramdisk) or stateful(ie. install to local >> disk). I know caoslinux can be installed to a local disk but can we >> centralize management of the local disk? >> >> Here is our ideal situation: >> i) new nodes are provisioned over the network (eg. pxeboot/tftp) >> ii) OS image/vnfs capsule is transferred over network and installed to >> local disk on the node as part of provisioning >> iii) a central "master" server/image is used for changes to a "master" >> OS image/vnfs capsule >> iv) changes to the master image are pushed out to the nodes (via rsync, >> xcpu, etc). We don't want to reboot a compute node to update the >> image/vnfs. We don't want a ramdisk for the compute node OS/vnfs install. >> >> My question is if caoslinux can handle stateful provisioning of a >> cluster. > > To quote a prominent vendor's recent keynote address: "In our > experience, stateful provisioning is never the right answer." > > That said, Caos is a distribution, not a cluster manager. It doesn't > care how it's provisioned. You can provision it statefully or > statelessly. > > Michael > > -- > Michael Jennings (a.k.a. KainX) http://www.kainx.org/ > Linux Server/Cluster Admin, LBL.gov Author, Eterm (www.eterm.org) > ----------------------------------------------------------------------- > "Sometimes I give myself the creeps. Sometimes my mind plays tricks > on me. It all keeps adding up; I think I'm cracking up. Am I just > paranoid? Am I just stoned?" -- Green Day, "Basket Case" > _______________________________________________ > Caos mailing list > Caos at lists.infiscale.org > http://lists.infiscale.org/mailman/listinfo/caos > From stefan at mdy.univie.ac.at Thu Nov 26 03:51:36 2009 From: stefan at mdy.univie.ac.at (Stefan Boresch) Date: Thu, 26 Nov 2009 12:51:36 +0100 Subject: [Caos] caos-testing: conflict abstractual-node vm-utils Message-ID: <20091126115136.GA13948@loop.mdy.univie.ac.at> Subject says it all: abstractual-node-1.1-0.20091117 and vm-utils-1.26-1 from caos-testing compete for /etc/sysconfig/vm and thus don't like each other ... Is vm-utils obsoleted by abstractual-node or vm-utils useful independently of abstractual ... Best regards, Stefan -- Stefan Boresch Institute for Computational Biological Chemistry University of Vienna, Waehringerstr. 17 A-1090 Vienna, Austria Phone: -43-1-427752715 Fax: -43-1-427752790 From stefan at mdy.univie.ac.at Thu Nov 26 05:12:09 2009 From: stefan at mdy.univie.ac.at (Stefan Boresch) Date: Thu, 26 Nov 2009 14:12:09 +0100 Subject: [Caos] ahci snafu of new livecd installer Message-ID: <20091126131209.GA13965@loop.mdy.univie.ac.at> Just got my new desktop (specs and output of caos sysinfo at the end of this mail) An install with SATA options in BIOS set to IDE (enhanced) works. However, it seems appropriate to switch to AHCI for a nice Intel SSD. So, I decided to switch SATA settings in BIOS from IDE enhanced to AHCI. The BIOS recognizes both SSD and DVD, however, the new installer does NOT find the CDROM anymore (very similar to my jmicron problems, which the Asus motherboard provides, btw, but which is out of the picture as the CD/DVD reader speaks SATA). Indeed, after using the preshell boot option and 'modprobe ahci' the install works without glitch. However, kernel panic upon reboot (no console found). Upon closer look, what is really going on is that the OS doesn't find the disk. (We are talking the 2.6.30.6-1 kernel at this point, but that is not the problem) So, I disect the initrd, and, surprise, no ahci/libata/scsi_mod/sd_mod modules are in the initrd :-( Just out of curiosity, I try the old installer (the one that initially installs Caos NSA 1.0.8, so I guess it's the one just before the liveCD). To my surprise, ahci gets detected automatically, and the system reboots just fine into the new kernel. Upon upgrading, first to 2.6.30.6-1, then to 2.6.31.6-2 from caos-testing, all works just fine. To sum up, I encountered 2 problems: (1) the liveCD installer fails to detect the need for ahci and friends. (2) Presumably because it was loaded by hand and not detected, ahci and supporting modules are not included in the initrd, resulting in a hung system upon first reboot. Let me repeat that installing with the old installer worked like a charmm, not even requiring 'preshell'. Does this mail suffice, or is a bug report the preferred way to go? Best regards, Stefan PS: Happy Thanksgiving! PPS: Machine specs: ---- Asus P6TSE, Core i7 920 Intel X25M Postville 80G SSD DVD-+R/-+RW SATA LG GH-22NS50 ASUS NVIDIA GTX260 card ---- caos sysinfo Main System: * 8 CPU(s): Intel(R) Core(TM) i7 CPU 920 @ 2.67GHz @ 1600.000 MHz * Memory total/free: 5970 MB / 5554 MB * Hostname: localhost @ 127.0.0.1 * Kernel: 2.6.31.6-2.caos #1 SMP Tue Nov 17 11:46:26 PST 2009 x86_64 Peripherial hardware: * Firewire (1106:3403): ZyAir G-110 802.11g * Graphics card (10de:5e2): * Network controller (r8169): RTL8111/8168B PCI Express Gigabit Ethernet controller * SCSI controller (ahci): * Sound card (snd-hda-intel): Network configuration: * lo: 127.0.0.1/255.0.0.0 * eth0: (r8169) 131.130.40.152/255.255.255.0 * Attached SCSI storage devices: Disk sda: size 74 GB, model: ATA INTEL SSDSA2M080 Disk sdb: size 0 GB, model: Generic USB SD Reader Disk sdc: size 0 GB, model: Generic USB CF Reader Disk sdd: size 0 GB, model: Generic USB xD/SM Reader Disk sde: size 0 GB, model: Generic USB MS Reader ---- -- Stefan Boresch Institute for Computational Biological Chemistry University of Vienna, Waehringerstr. 17 A-1090 Vienna, Austria Phone: -43-1-427752715 Fax: -43-1-427752790 From stefan at mdy.univie.ac.at Thu Nov 26 05:20:38 2009 From: stefan at mdy.univie.ac.at (Stefan Boresch) Date: Thu, 26 Nov 2009 14:20:38 +0100 Subject: [Caos] sidekick openoffice installation broken Message-ID: <20091126132038.GA14289@loop.mdy.univie.ac.at> the 3rd party option in sidekick for openoffice is broken as it looks for OO 3.0.0, which is not available anymore. I installed ooffice by hand; annoyingly, for IntelX86-64, I cannot find tar.gz files without JRE ... well, diskspace is cheap ... Best regards, Stefan -- Stefan Boresch Institute for Computational Biological Chemistry University of Vienna, Waehringerstr. 17 A-1090 Vienna, Austria Phone: -43-1-427752715 Fax: -43-1-427752790 From stefan at mdy.univie.ac.at Thu Nov 26 08:28:26 2009 From: stefan at mdy.univie.ac.at (Stefan Boresch) Date: Thu, 26 Nov 2009 17:28:26 +0100 Subject: [Caos] Mixing i386 and x86_64 packages ?? Message-ID: <20091126162826.GA22698@loop.mdy.univie.ac.at> OK, so maybe this is a stupid idea to begin with. However, I occasionally have the need to use acroread; I have encountered situtations where none of the free pdf viewers could do the job (I have also encountered pdfs that were perfectly rendered by xpdf or evince and of which acroread made a mess). So, I installed acroread straight from the adobe site only to find that it misses tons of i386 libs. I finally have it running after tediously searching for all the required rpms. Is there any chance that smart can do the work for me, i.e., smart install --some_option_indicating_that_I_want_i386 It seems like a RTFM question, but I seem to be blind at the moment ... It can obviously be done, because there are definitely i386 rpms on my (freshly installed system) which I did not install ;-) Thanks in advance, Stefan -- Stefan Boresch Institute for Computational Biological Chemistry University of Vienna, Waehringerstr. 17 A-1090 Vienna, Austria Phone: -43-1-427752715 Fax: -43-1-427752790 From gmk at infiscale.org Thu Nov 26 11:20:12 2009 From: gmk at infiscale.org (Greg Kurtzer) Date: Thu, 26 Nov 2009 11:20:12 -0800 Subject: [Caos] ahci snafu of new livecd installer In-Reply-To: <20091126131209.GA13965@loop.mdy.univie.ac.at> References: <20091126131209.GA13965@loop.mdy.univie.ac.at> Message-ID: <20091126192012.GE31788@infiscale.org> Hi Stefan, Can you test some "testing" ISO's that I have at: http://altruistic.infiscale.org/~gmk/isos/ I am not sure why ahci wasn't automatically found. Once booted, can you run the command: /sbin/detect -ai and send me the output? Thanks! Greg On Thursday, 26 November 2009, at 14:12:09 (+0100), Stefan Boresch wrote: > Just got my new desktop (specs and output of caos sysinfo at the end of > this mail) > > An install with SATA options in BIOS set to IDE (enhanced) works. > However, it seems appropriate to switch to AHCI for a nice Intel SSD. > So, I decided to switch SATA settings in BIOS from IDE enhanced to AHCI. > The BIOS recognizes both SSD and DVD, however, the new installer does NOT > find the CDROM anymore (very similar to my jmicron problems, which the > Asus motherboard provides, btw, but which is out of the picture as the > CD/DVD reader speaks SATA). > > Indeed, after using the preshell boot option and 'modprobe ahci' the > install works without glitch. However, kernel panic upon reboot (no > console found). Upon closer look, what is really going on is that the > OS doesn't find the disk. (We are talking the 2.6.30.6-1 kernel at > this point, but that is not the problem) > > So, I disect the initrd, and, surprise, no ahci/libata/scsi_mod/sd_mod modules > are in the initrd :-( > > Just out of curiosity, I try the old installer (the one that initially installs > Caos NSA 1.0.8, so I guess it's the one just before the liveCD). To my > surprise, ahci gets detected automatically, and the system reboots just > fine into the new kernel. Upon upgrading, first to 2.6.30.6-1, then to > 2.6.31.6-2 from caos-testing, all works just fine. > > To sum up, I encountered 2 problems: (1) the liveCD installer fails to > detect the need for ahci and friends. (2) Presumably because it was loaded > by hand and not detected, ahci and supporting modules are not included in > the initrd, resulting in a hung system upon first reboot. Let me repeat that > installing with the old installer worked like a charmm, not even requiring > 'preshell'. > > Does this mail suffice, or is a bug report the preferred way to go? > > Best regards, > > Stefan > > PS: Happy Thanksgiving! > PPS: Machine specs: > > ---- > > Asus P6TSE, Core i7 920 > Intel X25M Postville 80G SSD > DVD-+R/-+RW SATA LG GH-22NS50 > ASUS NVIDIA GTX260 card > > ---- > > caos sysinfo > Main System: > * 8 CPU(s): Intel(R) Core(TM) i7 CPU 920 @ 2.67GHz @ 1600.000 MHz > * Memory total/free: 5970 MB / 5554 MB > * Hostname: localhost @ 127.0.0.1 > * Kernel: 2.6.31.6-2.caos #1 SMP Tue Nov 17 11:46:26 PST 2009 x86_64 > > Peripherial hardware: > * Firewire (1106:3403): ZyAir G-110 802.11g > * Graphics card (10de:5e2): > * Network controller (r8169): RTL8111/8168B PCI Express Gigabit Ethernet controller > * SCSI controller (ahci): > * Sound card (snd-hda-intel): > > Network configuration: > * lo: 127.0.0.1/255.0.0.0 > * eth0: (r8169) 131.130.40.152/255.255.255.0 > > * Attached SCSI storage devices: > Disk sda: size 74 GB, model: ATA INTEL SSDSA2M080 > Disk sdb: size 0 GB, model: Generic USB SD Reader > Disk sdc: size 0 GB, model: Generic USB CF Reader > Disk sdd: size 0 GB, model: Generic USB xD/SM Reader > Disk sde: size 0 GB, model: Generic USB MS Reader > > ---- > > -- > Stefan Boresch > Institute for Computational Biological Chemistry > University of Vienna, Waehringerstr. 17 A-1090 Vienna, Austria > Phone: -43-1-427752715 Fax: -43-1-427752790 > _______________________________________________ > Caos mailing list > Caos at lists.infiscale.org > http://lists.infiscale.org/mailman/listinfo/caos -- Greg M. Kurtzer Chief Technology Officer HPC Systems Architect Infiscale, Inc. - http://www.infiscale.com From gmk at infiscale.org Thu Nov 26 11:21:28 2009 From: gmk at infiscale.org (Greg Kurtzer) Date: Thu, 26 Nov 2009 11:21:28 -0800 Subject: [Caos] caos-testing: conflict abstractual-node vm-utils In-Reply-To: <20091126115136.GA13948@loop.mdy.univie.ac.at> References: <20091126115136.GA13948@loop.mdy.univie.ac.at> Message-ID: <20091126192127.GF31788@infiscale.org> You are right, abstractual should obsolete vm-utils. Thanks for pointing that out! Greg On Thursday, 26 November 2009, at 12:51:36 (+0100), Stefan Boresch wrote: > Subject says it all: abstractual-node-1.1-0.20091117 and > vm-utils-1.26-1 from caos-testing compete for /etc/sysconfig/vm > and thus don't like each other ... > > Is vm-utils obsoleted by abstractual-node or vm-utils useful independently > of abstractual ... > > Best regards, > > Stefan > > -- > Stefan Boresch > Institute for Computational Biological Chemistry > University of Vienna, Waehringerstr. 17 A-1090 Vienna, Austria > Phone: -43-1-427752715 Fax: -43-1-427752790 > _______________________________________________ > Caos mailing list > Caos at lists.infiscale.org > http://lists.infiscale.org/mailman/listinfo/caos -- Greg M. Kurtzer Chief Technology Officer HPC Systems Architect Infiscale, Inc. - http://www.infiscale.com From stefan at mdy.univie.ac.at Thu Nov 26 23:35:24 2009 From: stefan at mdy.univie.ac.at (Stefan Boresch) Date: Fri, 27 Nov 2009 08:35:24 +0100 Subject: [Caos] ahci snafu of new livecd installer In-Reply-To: <20091126192012.GE31788@infiscale.org> References: <20091126131209.GA13965@loop.mdy.univie.ac.at> <20091126192012.GE31788@infiscale.org> Message-ID: <20091127073524.GC22698@loop.mdy.univie.ac.at> Hi Greg, thanks for the reply! [shouldn't you be celebrating Thanksgiving?] On Thu, Nov 26, 2009 at 11:20:12AM -0800, Greg Kurtzer wrote: > Can you test some "testing" ISO's that I have at: > > http://altruistic.infiscale.org/~gmk/isos/ > > I am not sure why ahci wasn't automatically found. Once booted, can you run > the command: /sbin/detect -ai and send me the output? > Not sure that I'll get around today, and Mon/Tue we have to organize a 'dog and pony show' as the department/faculty is being evaluated. But I'll get back to you. Thanks again, Stefan -- Stefan Boresch Institute for Computational Biological Chemistry University of Vienna, Waehringerstr. 17 A-1090 Vienna, Austria Phone: -43-1-427752715 Fax: -43-1-427752790 From stefan at mdy.univie.ac.at Fri Nov 27 00:58:27 2009 From: stefan at mdy.univie.ac.at (Stefan Boresch) Date: Fri, 27 Nov 2009 09:58:27 +0100 Subject: [Caos] ahci snafu of new livecd installer In-Reply-To: <20091127073524.GC22698@loop.mdy.univie.ac.at> References: <20091126131209.GA13965@loop.mdy.univie.ac.at> <20091126192012.GE31788@infiscale.org> <20091127073524.GC22698@loop.mdy.univie.ac.at> Message-ID: <20091127085827.GD22698@loop.mdy.univie.ac.at> Greg, On Fri, Nov 27, 2009 at 08:35:24AM +0100, Stefan Boresch wrote: > Not sure that I'll get around today, and Mon/Tue we have to organize > a 'dog and pony show' as the department/faculty is being evaluated. But > I'll get back to you. it may even take longer, since the machine just encountered massive file system corruption, i.e., the system is hosed. Thus, I'll have to weed out hardware issues first. [I noted that particularly during reboots / would not get unmounted / remounted ro correctly -- anyone having exp. with (Intel) SSD?] I just ran a detect -ai from the live CD just before desaster struck (without being able to save the output); one thing that was clear though: it literally recognizes nothing ...; in parcilar, no SCSI line if I remember correctly. Sorry about that, but I'll be in touch! Stefan -- Stefan Boresch Institute for Computational Biological Chemistry University of Vienna, Waehringerstr. 17 A-1090 Vienna, Austria Phone: -43-1-427752715 Fax: -43-1-427752790 From larnolda1 at gmail.com Sun Nov 29 15:46:18 2009 From: larnolda1 at gmail.com (Lou Arnold) Date: Sun, 29 Nov 2009 18:46:18 -0500 Subject: [Caos] CAOS Linux LiveDVD Installation Problem Message-ID: I downloaded the ISO, burned it, stuck it in my system and it merrily "tuned" my data drive and left the hard drive, with another OS on it, alone. I didn't lose much data, but it would be good to warn people in the installation instructions in the wiki, or change the installation process to let the user select what drive. I assume I will have to disconnect my data hard drive before I try installing again? -- Lou. From stefan at mdy.univie.ac.at Mon Nov 30 01:38:27 2009 From: stefan at mdy.univie.ac.at (Stefan Boresch) Date: Mon, 30 Nov 2009 10:38:27 +0100 Subject: [Caos] ahci snafu of new livecd installer Message-ID: <20091130093827.GB12477@loop.mdy.univie.ac.at> Hi Greg, downloaded the base CD (amd64) from http://altruistic.infiscale.org/~gmk/isos/ This one recognizes ahci (*and* jmicron, which is present on this system, but not used, so looks like this bug is closed -- if you want, I could test it on a machine where jmicron is needed). The output of detect -ai from the exp. livecd environment and from the regular livecd is attached. I will now install from the exp. livecd. However, I am still *scared* by the corrupted disk I encountered on the first attempt. I installed Ubuntu 9.10 (also a 2.6.31 kernel) on Friday and stresstested the machine (sys_basher, memtest86+). Within a few hours (I know that's short), no apparent problems with memory, nor with disk. Nevertheless, given that I didn't do anything stressful with Caos either, I doubt that there necessarily is a hardware problem (which was my first thought ...) Couple of things I noticed in this respect: (1) when restarting/shutting down, just before the screen/the machine shuts down, there are some SCSI / disk warnings / errors; occasionally during reboots at the end of the shutdown sequence I noted that the system doesn't remount / in ro only-mode. No clue whether that's a problem -- I never succeeded in reading these messages. (2) The Ubuntu 2.6.31 kernel "cheats" in the sense that (to my surprise) ahci, libata and friends, as well as ext3/ext4 are compiled into the kernel; no such modules are visible by lsmod and from the boot message (dmesg) one sees that ahci is loaded / available almost instantenously ... I wonder whether this explains why upon rebooting (starting up) I was very often thrown into a console and told to run fsck -A -T ... Best regards, Stefan PS: Can one establish a serial console when booting the install/LiveCD? This might be useful in the future ... PPS: Will do a simple base install for testing first; however, ultimately, I would like to have separate /boot, / and /scratch partitions (don't ask, long tradition here to have a "nonstandard" separate filesystem that hopefully is left alone by installers etc ...), and I would like to have / and /scratch as ext4 ... (/boot can remain whatever) Any words of wisdom / warning in this respect ?? ========================================================================== detect -ai from normal live cd installer (1.0.25) 8086:3405 BRIDGE unknown device 8086:3408 PCI unknown device 8086:340a PCI unknown device 8086:340e PCI unknown device 8086:342e OTHER unknown device 8086:3422 OTHER unknown device 8086:3423 OTHER unknown device 8086:3438 OTHER unknown device 8086:3a37 USB unknown device 8086:3a38 USB unknown device 8086:3a39 USB unknown device 8086:3a3c USB unknown device 8086:3a3e AUDIO snd-hda-intel 8086:3a40 PCI unknown device 8086:3a44 PCI unknown device 8086:3a46 PCI unknown device 8086:3a34 USB unknown device 8086:3a35 USB unknown device 8086:3a36 USB unknown device 8086:3a3a USB unknown device 8086:244e PCI unknown device 8086:3a16 BRIDGE iTCO_wdt 8086:3a22 SCSI unknown device 8086:3a30 SERIAL i2c-i801 10de:5e2 DISPLAY unknown device 1106:3403 IEEE1394 unknown device 10ec:8168 NETWORK r8169 ========================================================================== detect -ai from your exp. CD 8086:3405 BRIDGE unknown device 8086:3408 PCI unknown device 8086:340a PCI unknown device 8086:340e PCI unknown device 8086:342e OTHER unknown device 8086:3422 OTHER unknown device 8086:3423 OTHER unknown device 8086:3438 OTHER unknown device 8086:3a37 USB unknown device 8086:3a38 USB unknown device 8086:3a39 USB unknown device 8086:3a3c USB unknown device 8086:3a3e AUDIO snd-hda-intel 8086:3a40 PCI unknown device 8086:3a44 PCI unknown device 8086:3a46 PCI unknown device 8086:3a34 USB unknown device 8086:3a35 USB unknown device 8086:3a36 USB unknown device 8086:3a3a USB unknown device 8086:244e PCI unknown device 8086:3a16 BRIDGE iTCO_wdt 8086:3a22 SCSI ahci 8086:3a30 SERIAL i2c-i801 10de:5e2 DISPLAY unknown device 1106:3403 IEEE1394 unknown device 10ec:8168 NETWORK r8169 ========================================================================== -- Stefan Boresch Institute for Computational Biological Chemistry University of Vienna, Waehringerstr. 17 A-1090 Vienna, Austria Phone: -43-1-427752715 Fax: -43-1-427752790 From stefan at mdy.univie.ac.at Mon Nov 30 09:04:13 2009 From: stefan at mdy.univie.ac.at (Stefan Boresch) Date: Mon, 30 Nov 2009 18:04:13 +0100 Subject: [Caos] a strange time (hwclock?) phenomenon Message-ID: <20091130170413.GD12477@loop.mdy.univie.ac.at> As I mentioned in passing ("ahci snafu", which seems resolved in Greg's experimental ISO images), I have a freshly installed caos linux system where I occasionally get hung upon restart with Manual file system repair needed!" Login and run 'fsck -A -T' to resolve." (from the fsck -T -a / line right at the beginning of /etc/init.d/sysinit). I finally managed to observe the output from the fsck attempt that results in the above message (i.e., rc of fsck >= 4), and it complains about a 'time in the future' of the first / default superblock. The offset between dates (expected vs. that of the superblock) was suspiciously close to 1 hour. Thus, I am wondering whether the setting of / reading from hwclock causes problems somewhere. I noticed that Caos doesn't ask whether to use UTC or LOCAL (defaulting to LOCAL), but since Caos is the only OS on that machine, this shouldn't matter, right? I have now debug statements in /etc/init.d/sysconfig and I did indeed see one instance where upon startup hwclock gave a result which was an hour off; whereas hwclock --hctosys ended up with the right time (timezone is Europe / Vienna), and on that particular instance the first fsck -T -a passed fine. Reading /etc/init.d/sysinit and /etc/init.d/sysdown I don't see anything obviously fishy. Thus, I am wondering whether this rings a bell for anyone. Any hints appreciated!! Best regards, Stefan PS: One thing that I find strange is that when in graphical mode I can never read the output of the shutdown scripts. When I issue shutdown/reboot from the xfce menu, the screen blanks (but remains in graphics mode), the nvidia driver actually tries to restart X (???), sometimes even the slim login window appears. Only after a few seconds everything goes black an depending whether I chose shutdown or reboot, the system turns itself off or reboots. The fsck problems seem to occur most often when I restart as a normal user ... -- Stefan Boresch Institute for Computational Biological Chemistry University of Vienna, Waehringerstr. 17 A-1090 Vienna, Austria Phone: -43-1-427752715 Fax: -43-1-427752790 From mej at caoslinux.org Mon Nov 30 09:30:13 2009 From: mej at caoslinux.org (Michael Jennings) Date: Mon, 30 Nov 2009 09:30:13 -0800 Subject: [Caos] a strange time (hwclock?) phenomenon In-Reply-To: <20091130170413.GD12477@loop.mdy.univie.ac.at> References: <20091130170413.GD12477@loop.mdy.univie.ac.at> Message-ID: <20091130173013.GB22028@kainx.org> On Monday, 30 November 2009, at 18:04:13 (+0100), Stefan Boresch wrote: > As I mentioned in passing ("ahci snafu", which seems resolved in Greg's > experimental ISO images), I have a freshly installed caos linux > system where I occasionally get hung upon restart > > with > > Manual file system repair needed!" > Login and run 'fsck -A -T' to resolve." > > (from the fsck -T -a / line right at the beginning of > /etc/init.d/sysinit). I finally managed to observe the output from > the fsck attempt that results in the above message (i.e., rc of fsck > >= 4), and it complains about a 'time in the future' of the first / > default superblock. The offset between dates (expected vs. that of the > superblock) was suspiciously close to 1 hour. If you keep your hardware clock in local time instead of UTC, this discrepancy will appear on boot. A fix for this has been committed and is currently in the "testing" repo. Michael -- Michael Jennings (a.k.a. KainX) http://www.kainx.org/ Linux Server/Cluster Admin, LBL.gov Author, Eterm (www.eterm.org) ----------------------------------------------------------------------- "Morality cannot be legislated, but behaviour can be regulated. Judicial decrees may not change the heart, but they can restrain the heartless." -- Dr. Martin Luther King, Jr. From slaton at berkeley.edu Mon Nov 30 19:13:05 2009 From: slaton at berkeley.edu (Slaton Lipscomb) Date: Mon, 30 Nov 2009 19:13:05 -0800 Subject: [Caos] caos2 to caos-nsa upgrade, various questions Message-ID: <89cd2f950911301913w286f1d04jf2c53c07122c8b83@mail.gmail.com> Hi guys, So I'm finally upgrading our cluster from caos2 & warewulf 2.6, to caos-nsa / perceus. Many thanks for all the hard work you guys have put into this. I've spent the last couple of days getting up to speed on the docs and listserv, and doing the initial master node setup. In the process I've racked up a few questions. (1) Is it possible to specify manual partitioning with the new live media installer, or must I roll my own installation media with a custom kickstart file? (2) I came across some nearly year old listserv messages suggesting an autofs package would be coming soon, but I don't see this in testing. Is this still planned? We make heavy use of autofs here. (3) The pdsh packages appears to be build without the genders module. Can it be used with the "-a" flag and a simple machine file, and if so where is pdsh looking for this file? I tried populating /etc/hosts.pdsh which worked with caos2, but this was unsuccessful. (4) I daily jump back and forth between Debian, RHEL and caos2, and frankly it's hard to keep track of the exact rpc service names each uses for tcp wrappers (hosts.allow/deny). Does caos-nsa want to see "rpc-" prepended here, i.e. rpc.mountd, rpc.nfsd, rpc.lockd, rpc.statd, rpc.idmapd, rpc.rquotad? thanks in advance. best slaton Slaton Lipscomb Nogales Lab, Howard Hughes Medical Institute From gmk at infiscale.org Mon Nov 30 22:25:39 2009 From: gmk at infiscale.org (Greg Kurtzer) Date: Mon, 30 Nov 2009 22:25:39 -0800 Subject: [Caos] caos2 to caos-nsa upgrade, various questions In-Reply-To: <89cd2f950911301913w286f1d04jf2c53c07122c8b83@mail.gmail.com> References: <89cd2f950911301913w286f1d04jf2c53c07122c8b83@mail.gmail.com> Message-ID: <20091201062539.GI31788@infiscale.org> On Monday, 30 November 2009, at 19:13:05 (-0800), Slaton Lipscomb wrote: > Hi guys, Hiya Slaton! > So I'm finally upgrading our cluster from caos2 & warewulf 2.6, to > caos-nsa / perceus. Many thanks for all the hard work you guys have > put into this. :) > I've spent the last couple of days getting up to speed on the docs and > listserv, and doing the initial master node setup. In the process I've > racked up a few questions. > > (1) Is it possible to specify manual partitioning with the new live > media installer, or must I roll my own installation media with a > custom kickstart file? You can boot into the LiveMedia installer, get a command prompt. From there you can either copy from the existing kickstart templates and modify to suit your needs, or create a new one. Once you look at the syntax you will see how to create additional partitions or modify the existing config. > (2) I came across some nearly year old listserv messages suggesting an > autofs package would be coming soon, but I don't see this in testing. > Is this still planned? We make heavy use of autofs here. Planned?... sure. ;) > (3) The pdsh packages appears to be build without the genders module. > Can it be used with the "-a" flag and a simple machine file, and if so > where is pdsh looking for this file? I tried populating > /etc/hosts.pdsh which worked with caos2, but this was unsuccessful. Newer pdsh versions don't work the same way with the -a option AFAIK. I just wrap an alias which included "-w n[0000-xxxx]" options. > (4) I daily jump back and forth between Debian, RHEL and caos2, and > frankly it's hard to keep track of the exact rpc service names each > uses for tcp wrappers (hosts.allow/deny). Does caos-nsa want to see > "rpc-" prepended here, i.e. rpc.mountd, rpc.nfsd, rpc.lockd, > rpc.statd, rpc.idmapd, rpc.rquotad? I believe so, but I don't use tcp_wrappers often enough to remember. Thanks, Greg > thanks in advance. > > best > slaton > > Slaton Lipscomb > Nogales Lab, Howard Hughes Medical Institute > _______________________________________________ > Caos mailing list > Caos at lists.infiscale.org > http://lists.infiscale.org/mailman/listinfo/caos -- Greg M. Kurtzer Chief Technology Officer HPC Systems Architect Infiscale, Inc. - http://www.infiscale.com