OT: OpenBSD NFS Performance
Jordan Geoghegan wrote: > On 11/17/18 10:53, Predrag Punosevac wrote: > > On Sat, Nov 17, 2018 at 01:35:05AM +0100, Willi Rauffer wrote: > > > >> Hello, > >> > >> we want to make one logical volume out of several physical volumes, > but there is no \ > >> LVM (Logical Volume Manager) in OpenBSD! > >> Will there be a LVM in OpenBSD in the future? > >> > >> Thanks...Willi Rauffer, UNOBank.org > > P.S. OpenBSD's NFSv3 server and client implementation is pretty slow > so > > that begs the question how you are going to access that data pool. > > > I have an OpenBSD 6.3 NFS server, and it is able to achieve gigabit line > > speed no problem. I've transferred hundreds of terrabytes through that > thing and it hasn't let me down once. Most of the NFS clients > connected to it are CentOS 7 machines, and after a bit of fiddling, > line speed was achieved without issue. I can believe that as the NFS read performance is primarily client-driven. > The OpenBSD NFS client does seem to be a a tad slow though, and much > fiddling was required to get anywhere close to line speed with it. As I already said NFS read performance is primarily client-driven. Setting the read-ahead (for example, mount_nfs -a 4) is the biggest performance driver for reads. Unsurprisingly OpenBSD defaults to -a 1. predrag@oko$ more /etc/fstab|grep nfs 192.168.3.2:/data/nfs/hammer nfs rw,noatime,-a=4 0 0 Most of what I know about the topics was initiated by this wonderful post of Matt Dillon https://marc.info/?l=openbsd-misc=146130062830832=2 I would be very interested to learn what you have done to get OpenBSD NFS client speed close to 1 Gigabit (although at work I only use 10 Gigabit or InfiniBand gear so even 1 Gigabit is only of interest for my home setup). Cheers, Predrag P.S. Just for the record I would much rather see WAPBL ported and fully functional on OpenBSD than NFS performance improvment or even HAMMER2. WAPBL would actually make a real difference for my firewall/embedded OpenBSD deployments. HAMMER2 would be nice to have on my OpenBSD laptop but I can leave without it.
Error on 6.4 Errata page
On: https://www.openbsd.org/errata64.html line 9: https://www.openbsd.org/errata63.html;> should be: https://www.openbsd.org/errata64.html;>
A small newfs puzzle.
I'm on this # uname -a OpenBSD AngkorWat.rclayton.net 6.4 GENERIC.MP#364 amd64 # and I'm trying to write some file systems on this # disklabel -p g sd1 # /dev/rsd1c: type: SCSI disk: SCSI disk label: Rugged FW USB3 duid: 7e82b7f3472419e3 flags: bytes/sector: 512 sectors/track: 63 tracks/cylinder: 255 sectors/cylinder: 16065 cylinders: 243201 total sectors: 3907029168 # total bytes: 1863.0G boundstart: 0 boundend: 3907029168 drivedata: 0 16 partitions: #size offset fstype [fsize bsize cpg] c: 1863.0G0 unused i: 600.0G 128 4.2BSD 8192 65536 1 j: 600.0G 1258291328 4.2BSD 8192 65536 1 k: 663.0G 2516582528 4.2BSD 8192 65536 1 # so I do this # newfs sd1i /dev/rsd1i: 614399.9MB in 1258291072 sectors of 512 bytes 189 cylinder groups of 3264.88MB, 52238 blocks, 104960 inodes each super-block backups (for fsck -b #) at: 128, 6686592, 13373056, 20059520, 26745984, 33432448, 40118912, 46805376, 53491840, 60178304, 66864768, 73551232, 80237696, 86924160, 93610624, 100297088, 106983552, 113670016, 120356480, 127042944, 133729408, 140415872, 147102336, 153788800, 160475264, 167161728, 173848192, 180534656, 187221120, 193907584, 200594048, 207280512, 213966976, 220653440, 227339904, 234026368, 240712832, 247399296, 254085760, 260772224, 267458688, 274145152, 280831616, 287518080, 294204544, 300891008, 307577472, 314263936, 320950400, 327636864, 334323328, 341009792, 347696256, 354382720, 361069184, 367755648, 374442112, 381128576, 387815040, 394501504, 401187968, 407874432, 414560896, 421247360, 427933824, 434620288, 441306752, 447993216, 454679680, 461366144, 468052608, 474739072, 481425536, 488112000, 494798464, 501484928, 508171392, 514857856, 521544320, 528230784, 534917248, 541603712, 548290176, 554976640, 561663104, 568349568, 575036032, 581722496, 588408960, 595095424, 601781888, 608468352, 615154816, 621841280, 628527744, 635214208, 641900672, 648587136, 655273600, 661960064, 668646528, 675332992, 682019456, 688705920, 695392384, 702078848, 708765312, 715451776, 722138240, 728824704, 735511168, 742197632, 748884096, 755570560, 762257024, 768943488, 775629952, 782316416, 789002880, 795689344, 802375808, 809062272, 815748736, 822435200, 829121664, 835808128, 842494592, 849181056, 855867520, 862553984, 869240448, 875926912, 882613376, 889299840, 895986304, 902672768, 909359232, 916045696, 922732160, 929418624, 936105088, 942791552, 949478016, 956164480, 962850944, 969537408, 976223872, 982910336, 989596800, 996283264, 1002969728, 1009656192, 1016342656, 1023029120, 1029715584, 1036402048, 1043088512, 1049774976, 1056461440, 1063147904, 1069834368, 1076520832, 1083207296, 1089893760, 1096580224, 1103266688, 1109953152, 1116639616, 1123326080, 1130012544, 1136699008, 1143385472, 1150071936, 1156758400, 1163444864, 1170131328, 1176817792, 1183504256, 1190190720, 1196877184, 1203563648, 1210250112, 1216936576, 1223623040, 1230309504, 1236995968, 1243682432, 1250368896, 1257055360, newfs: ioctl (WDINFO): Input/output error newfs: /dev/rsd1i: can't rewrite disk label # and that doesn't look too cool, so I try this # fsck /dev/sd1i ** /dev/rsd1i ** File system is clean; not checking # Huh. But newfs is cheap, so I try it again # newfs sd1i /dev/rsd1i: 614399.9MB in 1258291072 sectors of 512 bytes 189 cylinder groups of 3264.88MB, 52238 blocks, 104960 inodes each super-block backups (for fsck -b #) at: 128, 6686592, 13373056, 20059520, 26745984, 33432448, 40118912, 46805376, 53491840, 60178304, 66864768, 73551232, 80237696, 86924160, 93610624, 100297088, 106983552, 113670016, 120356480, 127042944, 133729408, 140415872, 147102336, 153788800, 160475264, 167161728, 173848192, 180534656, 187221120, 193907584, 200594048, 207280512, 213966976, 220653440, 227339904, 234026368, 240712832, 247399296, 254085760, 260772224, 267458688, 274145152, 280831616, 287518080, 294204544, 300891008, 307577472, 314263936, 320950400, 327636864, 334323328, 341009792, 347696256, 354382720, 361069184, 367755648, 374442112, 381128576, 387815040, 394501504, 401187968, 407874432, 414560896, 421247360, 427933824, 434620288, 441306752, 447993216, 454679680, 461366144, 468052608, 474739072, 481425536, 488112000, 494798464, 501484928, 508171392, 514857856, 521544320, 528230784, 534917248, 541603712, 548290176, 554976640, 561663104, 568349568, 575036032, 581722496, 588408960, 595095424, 601781888, 608468352, 615154816, 621841280, 628527744, 635214208, 641900672,
Re: Missing LVM (Logical Volume Manager)
On 11/17/18 10:53, Predrag Punosevac wrote: On Sat, Nov 17, 2018 at 01:35:05AM +0100, Willi Rauffer wrote: Hello, we want to make one logical volume out of several physical volumes, but there is no \ LVM (Logical Volume Manager) in OpenBSD! Will there be a LVM in OpenBSD in the future? Thanks...Willi Rauffer, UNOBank.org P.S. OpenBSD's NFSv3 server and client implementation is pretty slow so that begs the question how you are going to access that data pool. I have an OpenBSD 6.3 NFS server, and it is able to achieve gigabit line speed no problem. I've transferred hundreds of terrabytes through that thing and it hasn't let me down once. Most of the NFS clients connected to it are CentOS 7 machines, and after a bit of fiddling, line speed was achieved without issue. The OpenBSD NFS client does seem to be a a tad slow though, and much fiddling was required to get anywhere close to line speed with it.
Re: OpenBSD migration
On Sat, Nov 17, 2018 at 10:42:57PM +0100, Ingo Schwarze wrote: > Hi Martin, > > Martin Sukany wrote on Sat, Nov 17, 2018 at 09:13:15PM +0100: > > > I want to migrate OpenBSD 6.4 (stable) from VM to bare metal. I see, as > > usual, two options: > > > > 1) install everything from scratch > > 2) create some flashimage (I did such thing on Solaris few years ago) > > and apply the image on new hw. > > I'd recommend option 1), reinstall. > > I have no idea whether or not option 2) will work. It may or may not. > If it doesn't, you end up doing a reinstall anyway, and nobody will be > interested in the reasons why it didn't work for you. Such a thing > simply isn't supported. > > Yours, > Ingo > I second reinstall. If you are concerned about setting things up the same I tend to do as couple tricks to make my setup portable. I have a tgz I created of important ~/.config items, mostly related to my openbox setup. I have git repos for my vim, mutt, fish, and ~/.local/bin items. I have a tgz of important /etc items, particularly vm.conf virtual host info for vmm, pf, etc. Lastly I have a master file of my installed packages created from pkg_info (forget the specific flags) but you can feel that file to a pkg add and let it install everything for you. Ken
Re: OpenBSD migration
Hi Martin, Martin Sukany wrote on Sat, Nov 17, 2018 at 09:13:15PM +0100: > I want to migrate OpenBSD 6.4 (stable) from VM to bare metal. I see, as > usual, two options: > > 1) install everything from scratch > 2) create some flashimage (I did such thing on Solaris few years ago) > and apply the image on new hw. I'd recommend option 1), reinstall. I have no idea whether or not option 2) will work. It may or may not. If it doesn't, you end up doing a reinstall anyway, and nobody will be interested in the reasons why it didn't work for you. Such a thing simply isn't supported. Yours, Ingo
OpenBSD migration
Hi, I want to migrate OpenBSD 6.4 (stable) from VM to bare metal. I see, as usual, two options: 1) install everything from scratch 2) create some flashimage (I did such thing on Solaris few years ago) and apply the image on new hw. I'd be glad for any personal experience / recommendations. NOTE: Server is not so important so downtime is not a problem here M>
Re: Missing LVM (Logical Volume Manager)
On 11/17/2018 10:53 AM, Predrag Punosevac wrote: On Sat, Nov 17, 2018 at 01:35:05AM +0100, Willi Rauffer wrote: Hello, we want to make one logical volume out of several physical volumes, but there is no \ LVM (Logical Volume Manager) in OpenBSD! Will there be a LVM in OpenBSD in the future? Thanks...Willi Rauffer, UNOBank.org There are people on this mailing list infinitely more knowledgeable and experienced than I both with Linux and BSDs so they will correct me claims if necessary. In my experience using LVM2 (LVM is depreciated) to create software RIAD even on Linux (I have the most experience with RHEL) is a bad idea unless you belive at the RedHat PR BS. Most people myself included if they have to use softraid on Linux prefer to do it from mdadm (softraid discipline for Linux and then perhaps put LVM on the top of it although I fail to see the purpose). In the lieu of the lack of modern file system on Linux (Btrfs is a vaporware and ZFS is an external kernel module which lags many version numbers behind Solaris and FreeBSD) some PR guys from RedHat started even advertising LVM2 snapshots as a real snapshots. That is pure BS as they are very expensive operation and for all practical purposes useless on the legacy file system XFS which is really the only really stable FS on Linux. If you are storing your data on Linux you should be using Hardware RAID and XFS. Not having LVM2 on OpenBSD is a feature not a bug! Dragon Fly BSD has partial not really functional implementation of LVM that I am quite familiar with. IIRC NetBSD has LVM2 implementation but it is hard to me to say usefulness of it as I have never used. As somebody mentioned. OpenBSD softraid can be used to manage logical volumes oko# bioctl softraid0 Volume Status Size Device softraid0 0 Online 2000396018176 sd3 RAID1 0 Online 2000396018176 0:0.0 noencl 1 Online 2000396018176 0:1.0 noencl but it is quite crude and it will take you more than a week to rebuild simple 10 TB mirror. IMHO softraid is far more useful for drive encryption on your laptop for example than for data storage. I don't have any experience with Hardware RAID cards on OpenBSD (Areca should have really good support) which I do prefer over softraid (but not over ZFS). However OpenBSD lacks modern file system (read HAMMER or HAMMER2) to take advantage of such set up. Best, Predrag P.S. OpenBSD's NFSv3 server and client implementation is pretty slow so that begs the question how you are going to access that data pool. I concur, software raid is a bug, not a feature, especially since if you truly need RAID, hardware cards are fairly cheap. But if you can't afford such a card, fairly reliable method is to just replicate the /altroot scheme with all your partitions. Even just using an external drive that you do periodic backups to is more reliable than software raid. For the most part, I've actually seen more failures with softraid than just independent disk even between systems where the only difference is the serial number being slightly incremented (sofraid, no matter how well coded still causes far more disk usage than a normal un-raided disk). Although, really, if you need reliability, it is much cheaper, less effort intensive, and more reliable to just grab a bunch of low-end systems and cluster them together. I have a small cluster 5 crusty old SunFire V120 boxes that've been running OpenBSD for nearly 10 years as my firewalls, I'm just running with a single disk in each. Each of them has failed at least a couple items over the years (failed disks, RAM modules, motherboards, power supplies, etc), but collectively they've had 100% reliability, even counting time for required reboots for upgrades, patches, and other maintenance Overall, I've found that software raid systems are only good for supporting whole-disk crypto and nothing else. Otherwise you are just adding an unnecessary performance penalty, kills your disks faster, and makes it much more a pain in the ass to recover from. -C .
Re: Missing LVM (Logical Volume Manager)
On Sat, Nov 17, 2018 at 01:35:05AM +0100, Willi Rauffer wrote: > Hello, > > we want to make one logical volume out of several physical volumes, but there > is no \ > LVM (Logical Volume Manager) in OpenBSD! > Will there be a LVM in OpenBSD in the future? > > Thanks...Willi Rauffer, UNOBank.org There are people on this mailing list infinitely more knowledgeable and experienced than I both with Linux and BSDs so they will correct me claims if necessary. In my experience using LVM2 (LVM is depreciated) to create software RIAD even on Linux (I have the most experience with RHEL) is a bad idea unless you belive at the RedHat PR BS. Most people myself included if they have to use softraid on Linux prefer to do it from mdadm (softraid discipline for Linux and then perhaps put LVM on the top of it although I fail to see the purpose). In the lieu of the lack of modern file system on Linux (Btrfs is a vaporware and ZFS is an external kernel module which lags many version numbers behind Solaris and FreeBSD) some PR guys from RedHat started even advertising LVM2 snapshots as a real snapshots. That is pure BS as they are very expensive operation and for all practical purposes useless on the legacy file system XFS which is really the only really stable FS on Linux. If you are storing your data on Linux you should be using Hardware RAID and XFS. Not having LVM2 on OpenBSD is a feature not a bug! Dragon Fly BSD has partial not really functional implementation of LVM that I am quite familiar with. IIRC NetBSD has LVM2 implementation but it is hard to me to say usefulness of it as I have never used. As somebody mentioned. OpenBSD softraid can be used to manage logical volumes oko# bioctl softraid0 Volume Status Size Device softraid0 0 Online 2000396018176 sd3 RAID1 0 Online 2000396018176 0:0.0 noencl 1 Online 2000396018176 0:1.0 noencl but it is quite crude and it will take you more than a week to rebuild simple 10 TB mirror. IMHO softraid is far more useful for drive encryption on your laptop for example than for data storage. I don't have any experience with Hardware RAID cards on OpenBSD (Areca should have really good support) which I do prefer over softraid (but not over ZFS). However OpenBSD lacks modern file system (read HAMMER or HAMMER2) to take advantage of such set up. Best, Predrag P.S. OpenBSD's NFSv3 server and client implementation is pretty slow so that begs the question how you are going to access that data pool.
Re: File sets on internet exposed server
Thank you Robert and Stuart for your helpful responses. > Skipping X and games is usually safe. The compilers might be a bad > idea unless you're only installing software from ports. Yes, current plan is to install only from ports as of now. > If you aren't using those packages which use libraries from xbase, you > *could* skip installing it, but make a note of it so that if you later > run pkg_add and get weird errors about missing libraries, you know what > you've done. Thanks for the detailed explanation. Will make a note. Most likely, the programs installed from ports should be fine. So, it is: -comp* -game* -x* Regards, ab -|-|-|-|-|-|-|--
Re: Missing LVM (Logical Volume Manager)
On Sat, Nov 17, 2018 at 01:35:05AM +0100, Willi Rauffer wrote: > Hello, > > we want to make one logical volume out of several physical volumes, but there > is no LVM (Logical Volume Manager) in OpenBSD! > > Will there be a LVM in OpenBSD in the future? > > Thanks...Willi Rauffer, UNOBank.org Probably not, but we have something that can do some of what LVM does: sofftraid. BTW, bugs@ is not the proper mailing list for this question. Redirected to misc@ -Otto