Re: [gentoo-user] best practise for 4 disks ... redundancy
Am 2015-03-13 um 15:11 schrieb Neil Bothwick: > On Fri, 13 Mar 2015 14:55:27 +0100, Stefan G. Weichinger wrote: > >>> See above. The reason for using a RAID1 array is to avoid >>> having to update multiple disks, just mount /boot on the RAID >>> device. >>> >>> % mount | grep boot /dev/md0 on /boot type ext2 >>> (rw,noatime,stripe=4) >> >> /boot is my vfat ESP here .. no ext2 ... > > That makes no difference, the filesystem sits on top of the RAID. > > As I said, that computer is pre-UEFI, so /boot is ext2. But RAID > doesn't care about the filesystem, it just mirrors the bits. Tried your suggestion today .. worked out great after adding that new array to dracut.conf ;) Thanks! goes into my howtos ...
Re: [gentoo-user] best practise for 4 disks ... redundancy
On Fri, 13 Mar 2015 14:55:27 +0100, Stefan G. Weichinger wrote: > > See above. The reason for using a RAID1 array is to avoid having > > to update multiple disks, just mount /boot on the RAID device. > > > > % mount | grep boot /dev/md0 on /boot type ext2 > > (rw,noatime,stripe=4) > > /boot is my vfat ESP here .. no ext2 ... That makes no difference, the filesystem sits on top of the RAID. As I said, that computer is pre-UEFI, so /boot is ext2. But RAID doesn't care about the filesystem, it just mirrors the bits. -- Neil Bothwick Pedestrians come in two types: Quick or Dead. pgp_XFPyDxuQP.pgp Description: OpenPGP digital signature
Re: [gentoo-user] best practise for 4 disks ... redundancy
Am 2015-03-13 um 09:59 schrieb Neil Bothwick: > See above. The reason for using a RAID1 array is to avoid having > to update multiple disks, just mount /boot on the RAID device. > > % mount | grep boot /dev/md0 on /boot type ext2 > (rw,noatime,stripe=4) /boot is my vfat ESP here .. no ext2 ...
Re: [gentoo-user] best practise for 4 disks ... redundancy
On Fri, 13 Mar 2015 09:38:15 +0100, Stefan G. Weichinger wrote: > c) Create a 4 disk RAID1 array consisting of sda2 and 3 > missing disks. Then add sdb2, sdc2, sdd2 in turn and the > contents of sda2 will be replicated to them. > >>> > >>> the array formatted as vfat (-> ESP) ? > > > >> sda2 is already formatted as FAT with the correct contents. > > yes ... and after that? > > install gummiboot to each physical disk? > create new loader-entries ? See above. The reason for using a RAID1 array is to avoid having to update multiple disks, just mount /boot on the RAID device. % mount | grep boot /dev/md0 on /boot type ext2 (rw,noatime,stripe=4) -- Neil Bothwick Ifyoucanreadthis,youspendtoomuchtimefiguringouttaglines. pgpPfrMAGwWSV.pgp Description: OpenPGP digital signature
Re: [gentoo-user] best practise for 4 disks ... redundancy
-BEGIN PGP SIGNED MESSAGE- Hash: SHA256 On 12.03.2015 19:52, Stefan G. Weichinger wrote: > On 12.03.2015 18:28, Neil Bothwick wrote: >> On Thu, 12 Mar 2015 18:12:06 +0100, Stefan G. Weichinger wrote: > c) Create a 4 disk RAID1 array consisting of sda2 and 3 missing disks. Then add sdb2, sdc2, sdd2 in turn and the contents of sda2 will be replicated to them. >>> >>> the array formatted as vfat (-> ESP) ? > >> sda2 is already formatted as FAT with the correct contents. yes ... and after that? install gummiboot to each physical disk? create new loader-entries ? -BEGIN PGP SIGNATURE- Version: GnuPG v2 iQIcBAEBCAAGBQJVAqH3AAoJEClcuD1V0Pzmz+MQAK8kAykLVEI+DC7tzf7UgF8/ M+yw56oL7UDFfv9t33KtwBFCAoG0hijag14zDm3/a/E0+LMZ1n2niyNJXvaqLzI5 ty2iPvfcnhVBnaydb1BZ2JRhtYEjs5jzVCFGpuKVN2ntBqoPdzYTZtJLtfSLS8Zz kyHHMLH6nQ2mjRKEP4HbDglHXCCU4y8jJMiB4Y/a4V6FFyaYvE6G0EuU9iZ6vfzY kvqE63UhLwvN+jNcAK/Ik7FHirQd+olvx1IFtVDQxFyMVHrVBfrhG/SSlmYs3lUy ftoFh5tFvYcy3z3aPss3oCwWH12330Y2EsAdsrZwliPawRfGDggb0wd8VDbc9Dod qzJUEjc5g0ih90fGJdAWHLumj+v4rl3aOq8zpF1uPjbG6DnbjbIsa2hRxmTf8nPS ASoDaU+NI4a62aJiwrUqAe4eEaSIBF40Kfd6Wz7wio85x3Yrkm2UVTJUtQTb2+3i VFOZrx1Khj/LkypyDqgv1BJYgvGwKn7zMrBjkKwKtBmcPZvo17ADwzsT+Eqj1SUP oycPdbtWhSrrjfzNbwcusnyMONhP7PNT7Bn9PRxnZM+alYqBghRN5lQkCfIA7FHC 5sC4e4x1qKV6JQlr25zqr7LnGVl3a4Kq+vX0mZ1E0isvgvzBs+pkkIdUjc78ShQz zwO6ycy74XoIpgV0bWt0 =0qGa -END PGP SIGNATURE-
Re: [gentoo-user] best practise for 4 disks ... redundancy
-BEGIN PGP SIGNED MESSAGE- Hash: SHA256 On 12.03.2015 18:28, Neil Bothwick wrote: > On Thu, 12 Mar 2015 18:12:06 +0100, Stefan G. Weichinger wrote: > >>> c) Create a 4 disk RAID1 array consisting of sda2 and 3 >>> missing disks. Then add sdb2, sdc2, sdd2 in turn and the >>> contents of sda2 will be replicated to them. >> >> the array formatted as vfat (-> ESP) ? > > sda2 is already formatted as FAT with the correct contents. partition type ? -BEGIN PGP SIGNATURE- Version: GnuPG v2 iQIcBAEBCAAGBQJVAeBhAAoJEClcuD1V0PzmneYQAKzbxmhfKWgS80u/CGQEYfqc hvR57LZ/z928N1H9hS25GYeneR8Wk2oEbXKCZQMzUbl5OwLi8wfK5cnbPxAPPAuj bx3IDnUdCM9xnFpeNymOlCNmBND+kR3ju365Ud198kuSyHBiMXfpUPQ+t0KsR6AP ilgDkRqABUf/5nBJQXMkBZeo5+ipMGKCCIUY+CNGJVfJ4yuoAYEnQiU5Y+CE2tl8 77h6NSerp+1xrAFu1NEwRS4ERjT4X+9hLDXoEa9lAfZn/7rrZdHTcHOLBo/X9aXU m7B82UbEHHY3Cdgef4Y4+Wfr25UpbqOHy6ZHzTYHJBgD72JJYZxZ2XizTUxdariS U424XSh0cXIlzW/Y2icX7ZnIWo5dvDBcKKOzhlI+LJAs1w1dzcLc4WcoFzku4Had WKDC/eVddnGtwaMzhqeoHZTK/mKLmZQJA2hwen2TcCQXZvEMjGGFfE1idcJkVOOQ PPXpNw71xM+ZymuFvpgaDhyj8BrX1sUey5fjVwUCANnr8/7CKyvcYAZ1C2+qRQ/c mNkbjc2KCAKjXTSGgk9Up51XQ+PtWE088l0WieOO5ayiosO523g2TVEjKLe2pE30 3SuAotpycVG5sRAzoHzzpBBaxCr3iAgujwyBstrYkjcMPJHZXwpSdCMm+IEZGCPi rzbR1AQQMoh11b/h1C6X =5BPh -END PGP SIGNATURE-
Re: [gentoo-user] best practise for 4 disks ... redundancy
On Thu, 12 Mar 2015 18:12:06 +0100, Stefan G. Weichinger wrote: > > c) Create a 4 disk RAID1 array consisting of sda2 and 3 missing > > disks. Then add sdb2, sdc2, sdd2 in turn and the contents of sda2 > > will be replicated to them. > > the array formatted as vfat (-> ESP) ? sda2 is already formatted as FAT with the correct contents. -- Neil Bothwick Tribble math: * + * = *** pgpdD57b8yJ9l.pgp Description: OpenPGP digital signature
Re: [gentoo-user] best practise for 4 disks ... redundancy
-BEGIN PGP SIGNED MESSAGE- Hash: SHA256 On 12.03.2015 17:48, Neil Bothwick wrote: > c) Create a 4 disk RAID1 array consisting of sda2 and 3 missing > disks. Then add sdb2, sdc2, sdd2 in turn and the contents of sda2 > will be replicated to them. the array formatted as vfat (-> ESP) ? will try, thanks -BEGIN PGP SIGNATURE- Version: GnuPG v2 iQIcBAEBCAAGBQJVAcjmAAoJEClcuD1V0PzmOH8QAKkne+w1gdU8oRvFSyQbYZd9 mJCM7z/Hf7QnBuibgfT12L1UtO8GNxnefElZ6mINHBlq1ULfGBW4RVwG/mUofjDA EYBJBh5O/olCqhhRL1wp287RD5GpCt/X+Q3Xv8uKsniiWtwzCaNqcRk23J0Ndsme IxOUzcg7LG0n1rRNJRCaUJoFqWmUXFthRfuf4dXzLsDCgbzp+l4JP941xZbjBM/W H5WZz+RLFlERi2ACS8+nNXlylDtc5rnzmPDg98iAjKWr+eS1OoQaZaEPcHS/ZnB8 tvyb+RlejS7IyV0ASpC9ExaEy5QGmqqkHrpaPpm2f8lvVoexRDJSHCLptxXcDuZM O83IM4Xl0Cl6Bsi8WSU4RQgvul30K701bgFYsWZkXXXKmpLtuEbd6UVQQ0e825VS NA3pWXbbpSARrDgHHBj97/WyZGsaRJ9GTIjbiZRjJd7y7g1reL99/rO8vXdfc1fx BRCIWM8PRvFjYaY0TeOybmBHALGOfZpuUU8qFkY9YG//9JnWYZer07QNtXn11JQF IkmdWlj/wc+sZHONUN9M0DMnXevhp0GzxpRvvv5o4dbxaQFhIQ2zHu3C4fHucqL4 /MDLDxJKTLY60u+82DYTCoafJYtmLf9nahpBx724Qg50dKb4BxTTpVc4oEO4L3S1 5/YOk7zSgPaW80E+SjAH =JO6L -END PGP SIGNATURE-
Re: [gentoo-user] best practise for 4 disks ... redundancy
On Thu, 12 Mar 2015 12:30:45 +0100, Stefan G. Weichinger wrote: > I now have a dedicated ESP on /dev/sda2 (non-RAID) and it boots via > gummiboot (-> UEFI). > > I now wonder if I could > > (a) create a RAID1 over /dev/sd[abcd]2 and format it as vfat ESP > > or > > (b) "clone" /dev/sda2 -> format each /dev/sd[abcd]2 as vfat, copy the > content of /boot in there and install gummiboot to each harddisk > c) Create a 4 disk RAID1 array consisting of sda2 and 3 missing disks. Then add sdb2, sdc2, sdd2 in turn and the contents of sda2 will be replicated to them. -- Neil Bothwick Make it idiot proof and someone will make a better idiot. pgpPAqvtBJEX9.pgp Description: OpenPGP digital signature
Re: [gentoo-user] best practise for 4 disks ... redundancy
-BEGIN PGP SIGNED MESSAGE- Hash: SHA256 On 11.03.2015 23:32, Neil Bothwick wrote: > On Wed, 11 Mar 2015 23:14:49 +0100, Stefan G. Weichinger wrote: > >> Am 11.03.2015 um 17:06 schrieb Neil Bothwick: >> So RAID1 with vfat as ESP ? >>> >>> That's how I'd do it. I have /boot on RAID1 on this box (but >>> not as FAT and it's a BIOS motherboard) with everything else on >>> a btrfs RAID and it works well. >> >> I will redo my setup tomorrow ... thanks So your suggestion is >> with MBR and not GPT? > > Absolutely not! MBR is fragile, limited and horrible. I use GPT but > not UEFI because this is a pre-UEFI motherboard. > >> I still wonder if to do UEFI and GPT because both are the more >> modern ways of doing things (and be prepared for future >> developments) or just stay with BIOS and MBR as long as I don't >> need it (with disks smaller than 2 TB etc) > > Given the choice, I'd use GPT for the above reasons and UEFI > because Gummiboot is nicer than GRUB. I now have a dedicated ESP on /dev/sda2 (non-RAID) and it boots via gummiboot (-> UEFI). dracut assembles the arrays after configuring this: # /etc/dracut.conf add_dracutmodules+="lvm bash mdraid" hostonly="yes" mdadmconf="yes" # the uuids of my two arrays ... kernel_cmdline+=" rd.md.uuid=7663e24a:1a64ce0f:49545629:46742031 " kernel_cmdline+=" rd.md.uuid=8c266e7a:870c97b7:ff199683:6d8c3edb " # end (lvm isn't necessary but doesn't hurt) I now wonder if I could (a) create a RAID1 over /dev/sd[abcd]2 and format it as vfat ESP or (b) "clone" /dev/sda2 -> format each /dev/sd[abcd]2 as vfat, copy the content of /boot in there and install gummiboot to each harddisk ? Stefan -BEGIN PGP SIGNATURE- Version: GnuPG v2 iQIcBAEBCAAGBQJVAXjlAAoJEClcuD1V0PzmVhIP/0B2ff9NNznkbkhMDtd850E3 lZKWIYSkEYgdznbPF0ODFV/dRZZHKjyAxgeEm1QbhJGRYCbOHMopcCMD2BQ+Cimt Fd0Eful1l2Z7dfr1igPCHHPJxgi65OvXd1KSFF9dSKPXaDe3B+QpG0QTBf+YU83p slAMmxwePXnKxFpZi2o7/7Nfi7SMDmDvJREgLL6BhFdDDQfVeMbsWjvtc2prutVM 858Eaou1hXkOc1d0Wkjhkyazw9VmpMyfN92+4YMQ+kWyz+x07WHE0gXMd+tvkTOw sOY7QYwKrsTXHdU5UTfHork+MVV6hJ2cLwj/epYXxU9T8zTEbaCkqlGoHbrGI1qF YCwGy89pkxVx5UZ+/1qSfalp99QiXed9fcDsrh8roHmUzSiQZOCoak9Wi+94S1cL 0Jc2vvVq9LUYAYiJ6/poUsZWB4W/LgKiB42YwFFwdoXMI+GIXWfmbB1B0mSnYYaU B+mtuv3trdgpD8+uyLjtmDOzbSBZekqG6RWHCdbh2R7mqmPoM3n8rmIz6R1+1pLy ynDJPot9hKxWaxdzu+SnvhlS/Kj04Qcydo2oav135HyoNJNN0j38E9BiRmYmBBfy CI1+MwR55tTriVgNHF8hEhH1eiqS+l6XE3ymuj+TLk9cKuR4lBtop+NG59r3ygTm miPw7o5+IPgI7bRcJiEV =ge3y -END PGP SIGNATURE-
Re: [gentoo-user] best practise for 4 disks ... redundancy
On Wed, 11 Mar 2015 23:14:49 +0100, Stefan G. Weichinger wrote: > Am 11.03.2015 um 17:06 schrieb Neil Bothwick: > > >> So RAID1 with vfat as ESP ? > > > > That's how I'd do it. I have /boot on RAID1 on this box (but not as > > FAT and it's a BIOS motherboard) with everything else on a btrfs > > RAID and it works well. > > I will redo my setup tomorrow ... thanks > So your suggestion is with MBR and not GPT? Absolutely not! MBR is fragile, limited and horrible. I use GPT but not UEFI because this is a pre-UEFI motherboard. > I still wonder if to do UEFI and GPT because both are the more modern > ways of doing things (and be prepared for future developments) or just > stay with BIOS and MBR as long as I don't need it (with disks smaller > than 2 TB etc) Given the choice, I'd use GPT for the above reasons and UEFI because Gummiboot is nicer than GRUB. -- Neil Bothwick During a raid on a local chemist's shop, 2000 Viagra tablets were stolen Police are looking for hardened criminals! pgp_BC1q0eDXr.pgp Description: OpenPGP digital signature
Re: [gentoo-user] best practise for 4 disks ... redundancy
-BEGIN PGP SIGNED MESSAGE- Hash: SHA256 Am 11.03.2015 um 17:06 schrieb Neil Bothwick: >> So RAID1 with vfat as ESP ? > > That's how I'd do it. I have /boot on RAID1 on this box (but not as > FAT and it's a BIOS motherboard) with everything else on a btrfs > RAID and it works well. I will redo my setup tomorrow ... thanks So your suggestion is with MBR and not GPT? I still wonder if to do UEFI and GPT because both are the more modern ways of doing things (and be prepared for future developments) or just stay with BIOS and MBR as long as I don't need it (with disks smaller than 2 TB etc) S -BEGIN PGP SIGNATURE- Version: GnuPG v2 iQIcBAEBCAAGBQJVAL5ZAAoJEClcuD1V0PzmLKoP/1dZwbyBmu4qbnQF8+IXKOHd QGeoGpoyP0ZYQaAoVNvklYNqeSYYsKHY0vV7olcAj28NLT6UsFd3J3NDbgxw9OlP KecuwIGdOylsPF+NDJ2hM5LlREVvhNzsHnMgkmzjaq6F+30rBTNIKSYJNxsxX26q wEjQIEPxFHxuafaW8tFROZqHtwr56XPf9PEB2DDrt2jK5H+roUx0u4BU6gLnQO3W Zf6tdbfW0rlRhllPGJUKaJOQp1sOnez7C2WToyVJX2Jy44pbuZ1SpUbRl008JY+6 M4QMSqbCvb41ud8TYgkZetFKHB9r/xLN0uE1oq1rlFh0mbw7TU+K+Lt4ICA/u/cN JVFM7tKuO+6wgzl+zJOWfbuGwXDAjYElDzquHCGIu823oezb6LXfA5GMJMqC2//e CnP4irR+D3fZIj5i5E6+7+R1wy2KVgm7FDinYmhHHzr1X/8Nj+ff1jGZx2j1gTk3 Mzc68GKYzewIJn5iR7/jt02FIvo0mVrkGCBMgnDnI90hFwUzBOCBdbkYp+j4kKRe d10/8mdlVwnCS0uXqXLCMbnaMYzZDn3YOrEdoNEkXkXxnUKOmfMqwTrT2TRvCyVW iz5r829IrXi80rtBn2+x9zTMZXodWUEj+Nzd2k2ok7hAqkIImcGXOAjh/Qj6wHbq xVNLmuNC7mnQTLWyrva5 =VI4K -END PGP SIGNATURE-
Re: [gentoo-user] best practise for 4 disks ... redundancy
On Wed, 11 Mar 2015 16:52:27 +0100, Stefan G. Weichinger wrote: > > I generally set up boot on a RAID1 across all disks. You can still > > access individual disks in a RAID1 array, so you can set your > > computer to boot from sda1 then sda2 etc but mount /boot on the > > array, so all updates are pushed to all four disks. That should have read "sda1 then sdb1". > So RAID1 with vfat as ESP ? That's how I'd do it. I have /boot on RAID1 on this box (but not as FAT and it's a BIOS motherboard) with everything else on a btrfs RAID and it works well. -- Neil Bothwick A: Because it messes up the order in which people normally read text. Q: Why is top-posting such a bad thing? A: Top-posting. Q: What is the most annoying thing on usenet and in e-mail? pgpbnAyzn5Dy2.pgp Description: OpenPGP digital signature
Re: [gentoo-user] best practise for 4 disks ... redundancy
-BEGIN PGP SIGNED MESSAGE- Hash: SHA256 On 11.03.2015 12:09, Neil Bothwick wrote: >> Then I spent some hours fiddling with grub2, gummiboot et al to >> get that box booting up fine. dracut doesn't assemble the >> rootfs-raid etc etc ... -> fun > > I found this with a LUKS root filesystem, dracut doesn't > automatically detect such devices any more, so speed up the boot. > You need to specify rd.auto or rd.mdraid to force it. I think, both ... I will have a look next time. > I generally set up boot on a RAID1 across all disks. You can still > access individual disks in a RAID1 array, so you can set your > computer to boot from sda1 then sda2 etc but mount /boot on the > array, so all updates are pushed to all four disks. So RAID1 with vfat as ESP ? S -BEGIN PGP SIGNATURE- Version: GnuPG v2 iQIcBAEBCAAGBQJVAGS7AAoJEClcuD1V0PzmbLMP/1aC6o90RCrVahLzkOo2tHj4 GeZ1Fvni//wcH/gc7GxxeTfkd0N6oY1O8LOcoA/zdHnKmJXuix52uunG7HGuu4Ti 99AntboNP+9zYn49Q47wodMaLFDtkFXaj22dBjUVX6dzHG4Fo3pe91NnXxwu4JXv HgaXUfpI8lWRmj+eFdWG6QjztQnPQug6JAt+6Lm1pcy59JmK6VC+YpAezBW2iyO6 Lz5YZtDga6YPrKHlIrhEN+KS0DW3vcenQQyNZPi41p/4B1LtD0A8EvCS6kOcj8eW t+JPGZXJ2JYA42aL/xIEuv8yTO5g2CP+X5vRPmPQvQBbDhmuBtgmoU/w/nVHczcO v8bywNGeYC+MjlOq4/WvCFXiwuBPcecx9DtyVVJKU+uqVpeV3n0mGxDaUY6yOP+j GfexRTBGVu/USspZI6h//+yUJ1Gn8NIQnr9CVscJc5aYz9sfIpUPmKjgdrL/QkpL u2TTu7Yo6T858ULt1oALbBFps3110wTu3g/czMp//5DIsLO8O0dQi6g7K6mg23J5 8vA5GKRZ46jM23avEe1+1aRva87e5/TElSpYF7VbA3+aYG7ZCNct7mgXHhnpO47n qBYV4CtI9gdaBY/teaSAo/JvfbwxyFMCvc9etj803JWPAgmOUwDOsrzfV77w5bti c3+I2Go9tGZ1BPwUC6UT =4li3 -END PGP SIGNATURE-
Re: [gentoo-user] best practise for 4 disks ... redundancy
On Wed, 11 Mar 2015 07:45:54 +0100, Stefan G. Weichinger wrote: > yesterday I received a shiny new server for a client. > > It brings 4 x 1 TB SATA disks, is capable of UEFI and I started setting > it up. > > GPT partitioning, small partition(s) for the ESP, some swap, and bigger > partitions of type fd=raid for creating 2 mdadm-raids (rootfs and data) > with level 6. > > ok so far. > > Then I spent some hours fiddling with grub2, gummiboot et al to get that > box booting up fine. dracut doesn't assemble the rootfs-raid etc etc ... > -> fun I found this with a LUKS root filesystem, dracut doesn't automatically detect such devices any more, so speed up the boot. You need to specify rd.auto or rd.mdraid to force it. > even if I manage to have it booting via UEFI from /dev/sda1 or so ... I > still have this single point of failure as the vfat-partitioned ESP is > on one physical disk only. > > I would like to set it up in a way that it boots even when the /dev/sda > dies ... I generally set up boot on a RAID1 across all disks. You can still access individual disks in a RAID1 array, so you can set your computer to boot from sda1 then sda2 etc but mount /boot on the array, so all updates are pushed to all four disks. -- Neil Bothwick I distinctly remember forgetting that. pgpXUdOVNwLSz.pgp Description: OpenPGP digital signature
[gentoo-user] best practise for 4 disks ... redundancy
greetings again yesterday I received a shiny new server for a client. It brings 4 x 1 TB SATA disks, is capable of UEFI and I started setting it up. GPT partitioning, small partition(s) for the ESP, some swap, and bigger partitions of type fd=raid for creating 2 mdadm-raids (rootfs and data) with level 6. ok so far. Then I spent some hours fiddling with grub2, gummiboot et al to get that box booting up fine. dracut doesn't assemble the rootfs-raid etc etc ... -> fun What the question is in general: even if I manage to have it booting via UEFI from /dev/sda1 or so ... I still have this single point of failure as the vfat-partitioned ESP is on one physical disk only. I would like to set it up in a way that it boots even when the /dev/sda dies ... Maybe I have to go the BIOS way and have an ext2-boot-raid1 over 4 disks/partitions? I did that on other servers back then. What is the up-to-date and recommended way of achieving this? How do you gentoo-users do these things? I am still at the start with this server and don't mind backing up the rootfs and start from scratch with partitioning ;) Thanks, Stefan