OpenBSD Journal

New flashboot release imminent, beta testers needed

Contributed by sean on from the no flash photography dept.

Rickard Dahlstrand writes:

A new version of flashboot is ready and we need people to betatest it before we put it up.

This release which is based on 4.0 stable includes among other things a new extension system with support for (most) standard OpenBSD packages.

As part of this release we also provide 128 mb diskimages.

Download kernels and images here http://tilde.se/flashboot/download/4.0/

Thanks, Rickard.


More information about flashboot can be found at it's website.

(Comments are closed)


Comments
  1. By Anonymous Coward (24.37.236.100) on

    Why not just do a 'full' -STABLE install and run in MFS, easy as 123...

    Just curious.

    Comments
    1. By Anonymous Coward (83.31.214.174) on

      takes 512MB CF since 3.8, I think.

      Comments
      1. By Craig B (220.233.217.230) on

        > takes 512MB CF since 3.8, I think.
        >

        Remove the DEBUG=-g flag - drops the build by quite a bit. (This was the reason for the big increase in lib size)

        Comments
        1. By Anonymous Coward (24.37.236.100) on

          > > takes 512MB CF since 3.8, I think.
          > >
          >
          > Remove the DEBUG=-g flag - drops the build by quite a bit. (This was the reason for the big increase in lib size)

          Remove this from where?

          Comments
          1. By Anonymous Coward (76.21.114.63) on

            > Remove this from where?

            man mk.conf

        2. By sthen (85.158.44.148) on

          > > takes 512MB CF since 3.8, I think.
          >
          > Remove the DEBUG=-g flag - drops the build by quite a bit.

          Why rebuild when you can strip?

          Comments
          1. By Anonymous Coward (85.178.103.217) on

            > > > takes 512MB CF since 3.8, I think.
            > >
            > > Remove the DEBUG=-g flag - drops the build by quite a bit.
            >
            > Why rebuild when you can strip?
            >

            man gzexe....

            Comments
            1. By sthen (85.158.44.148) on

              > > > > takes 512MB CF since 3.8, I think.
              > > > Remove the DEBUG=-g flag - drops the build by quite a bit.
              > > Why rebuild when you can strip?
              >
              > man gzexe....

              hmm, well I used to, but a gig CF costs what a 32mb did in those days and it just doesn't seem worth the hassle any more...

      2. By RC (71.105.35.36) on

        > takes 512MB CF since 3.8, I think.

        That's why I stripped it down and made a copy available: http://rc.xoompages.com/

        Only 28MBs, and a (damn-near) fully functional OpenBSD base system. You can add any of the binary packages you want, and add-on addition dist sets if, for some reason, you want to.

        It's really, really amazing how much irrelevent junk is in base.

        Comments
        1. By sthen (85.158.44.148) on

          > > takes 512MB CF since 3.8, I think.
          >
          > That's why I stripped it down and made a copy available: http://rc.xoompages.com/

          "I removed duplicate files (eg. gzip/gunzip) and made them tiny symlinks instead."

          if you ended up with gzip and gunzip being separate files (as opposed to separate directory entries), you've done something wrong...

      3. By henning (80.171.86.179) on

        > takes 512MB CF since 3.8, I think.

        no, 256 is still fine.

        <brahe@cr10.ams> $ uname -a
        OpenBSD cr10.ams.bsws.de 4.0 GENERIC#5 i386
        <brahe@cr10.ams> $ grep ^wd0 /var/run/dmesg.boot
        wd0 at pciide0 channel 0 drive 0: <SAMSUNG CF/ATA>
        wd0: 1-sector PIO, LBA, 248MB, 508928 sectors
        wd0(pciide0:0:0): using PIO mode 4
        <brahe@cr10.ams> $ df -h
        Filesystem Size Used Avail Capacity Mounted on
        /dev/wd0a 244M 182M 49.6M 79% /
        mfs:2505 121M 1.0K 115M 0% /tmp
        mfs:21279 752K 345K 369K 48% /dev
        mfs:8198 9.2M 3.4M 5.4M 39% /var

      4. By Lars Hansson (203.65.246.12) lars@unet.net.ph on

        > takes 512MB CF since 3.8, I think.

        base fits comfortably on 256Mb:
        pengo$ df -h
        Filesystem Size Used Avail Capacity Mounted on
        /dev/wd0a 239M 173M 54.9M 76% /
        mfs:4950 239K 33.0K 195K 14% /dev
        mfs:16117 7.7M 2.3M 5.0M 32% /var
        mfs:14206 3.9M 2.0K 3.7M 0% /tmp

        pengo$ uname -a
        OpenBSD pengo.unet.net.ph 3.9 GENERIC#1 i386

        pengo$ dmesg | grep wd0
        wd0 at pciide0 channel 0 drive 0: <Hitachi XXM2.3.0>
        wd0: 1-sector PIO, LBA, 244MB, 500400 sectors
        wd0(pciide0:0:0): using PIO mode 1
        dkcsum: wd0 matches BIOS drive 0x80
        root on wd0a

  2. By SleighBoy (64.95.191.180) on http://www.code.cx/

    I am currently trying a Solid-State IDE drive in a 4801 I setup. Using remote syslog and no mail to reduce writes. They say the drive is good for 6 million+ writes and comes with a 7 year warranty.

    http://www.flickr.com/photos/sleighboy/359004413/

    These drives are not cheap, but hopefully it will provide a more normal OS install without the longevity issues that a normal notebook IDE drive has.

    I'd like to hear anybodies comments on use such a drive, specifically one from this manufacturer.

    Comments
    1. By Anonymous Coward (193.10.185.3) on

      Has anyone actually managed to corrupt their compact flash card in a live system that doesnt swap? ie a router and a standard OS install.

      Comments
      1. By Leon Yendor (218.214.194.113) on

        > Has anyone actually managed to corrupt their compact flash card in a live system that doesnt swap? ie a router and a standard OS install.
        >
        >
        I have been running a Soekris 4801 for over a year on just one Apacer 512MB CF.

        It has had a normal install of OpenBSD 3.8 and 3.9 and is due for 4.0 shortly. Until Xmas it was running spamd with verbose logging guarding a mailserver on the server LAN.

        With 1GB CF cards around I'd expect the spare cell array to be able to handle heaps more and they are now lower price than the 256MB units I started with a couple of years ago.

        I have NEVER had a failure and $deity knows I have tried to overwork some of them. I have used Sandisk and Apacer PhotoSteno II and III and found that they Just Work (tm) though the Apacers seem faster particularly the 88X III's.

        It is nice to just do a PXE boot to my install server, put on the sets I need (no comp, games or X but man always goes on) copy or merge the configs as needed and away they go for another six months. KISS works - who needs complexity?

      2. By Peter Hopfgartner (217.199.20.139) on http://www.hopfgartner.it

        I had a corrupted flash card on s Soekris 4501 last week. It carried a OBSD 3.9 installation used as a router/firewall/openvpn gateway. The card was a kingston 512 MB (not one of those secure/industrial grade/etc cards).

        Regards,

        Peter

    2. By Uwe Hollerbach (66.168.114.99) on

      > I am currently trying a Solid-State IDE drive in a 4801 I setup. Using remote syslog and no mail to reduce writes. They say the drive is good for 6 million+ writes and comes with a 7 year warranty.
      >
      > http://www.flickr.com/photos/sleighboy/359004413/
      >
      > These drives are not cheap, but hopefully it will provide a more normal OS install without the longevity issues that a normal notebook IDE drive has.
      >
      > I'd like to hear anybodies comments on use such a drive, specifically one from this manufacturer.

      I dunno, I've not yet had any (lack of) longevity issues with a plain old notebook IDE drive; I've had a 4801 as a firewall + mail server, with a completely normal OS install (no swap, and it's true I set noatime on all partitions on that drive), for several years now -- 3.5ish? The original drive was a cast-off, having been used previously for a couple of years in an actual laptop, and it never had any issues. I did replace it, but only because I was starting to get just a bit tight on space; this was a 1 GB drive. Now I have a 60 GB drive, the smallest I could find, and it too has been running 24x7 for a good long time.

      Out of curiosity, how expensive was this solid-state drive? I had a look at the manufacturer's web site, but they don't seem to say there. But my guess would be that even if regular old spinning platters fail, it would still be cheaper to replace one every couple of years than to get one of these. Of course, if this is going to be placed on top of a tower in 40-below weather 400 miles from the nearest road, it's probably worth going for the more rugged solution...

      UH

      Comments
      1. By SleighBoy (24.113.29.23) on http://www.code.cx/

        > Out of curiosity, how expensive was this solid-state drive? I had a look > at the manufacturer's web site, but they don't seem to say there.

        They are for sale at mouser.com

        This particular one was a 2GB, and was $308.00 USD

        Comments
        1. By Mark Carlson (dextromulous) on http://mcarlson.ca

          > > Out of curiosity, how expensive was this solid-state drive? I had a look > at the manufacturer's web site, but they don't seem to say there.
          >
          > They are for sale at mouser.com
          >
          > This particular one was a 2GB, and was $308.00 USD
          >

          There are also some less rugged ones on Digikey (M-Systems / SanDisk,) 1GB for $93 CAD (2.5")

          The performance would still probably be good enough for most people.

  3. By Rickard Dahlstrand (213.113.219.180) on

    I for one use flashboot on 50+ devices and like the fact that I can upgrade them all by scp, and keep multiple kernels (full dist) on flash at once. Most of my systems are really simple and doesn't need much from the normal install, so the added functionality doesn't do much.

    The ramdisk also adds speed and reliability and with the flash out of the way there are not really any components that will fail in a 5 to 10-year period and that saves time and money having to go to all these places to replace broken devices.

    While using flashdist (a really great system) a long time ago I had devices fail on me because of flash-wear. I now use industrial grade flash cards so this might have been a none issue but still, better safe.. Libc was the one that failed for me since I ran a shellscript (including a couple of binaries) once every minute for about 7 months.

    So, flashboot is the only working solution for me. It might not suit everyone. But if you have a large number of devices that you want to upgrade easily, are worrying about flash-wear and doesn't mind the limited functionality then flashboot could be working for you as well.

    And hopefully with the release of diskimages we have made it much easier for other people to install it and try it out. (Using dd or physdiskwrite on Windows)

    Rickard.

    Comments
    1. By Anonymous Coward (24.37.236.100) on

      > I for one use flashboot on 50+ devices and like the fact that I can upgrade them all by scp, and keep multiple kernels (full dist) on flash at once. Most of my systems are really simple and doesn't need much from the normal install, so the added functionality doesn't do much.
      >
      > The ramdisk also adds speed and reliability and with the flash out of the way there are not really any components that will fail in a 5 to 10-year period and that saves time and money having to go to all these places to replace broken devices.
      >
      > While using flashdist (a really great system) a long time ago I had devices fail on me because of flash-wear. I now use industrial grade flash cards so this might have been a none issue but still, better safe.. Libc was the one that failed for me since I ran a shellscript (including a couple of binaries) once every minute for about 7 months.
      >
      > So, flashboot is the only working solution for me. It might not suit everyone. But if you have a large number of devices that you want to upgrade easily, are worrying about flash-wear and doesn't mind the limited functionality then flashboot could be working for you as well.
      >
      > And hopefully with the release of diskimages we have made it much easier for other people to install it and try it out. (Using dd or physdiskwrite on Windows)
      >
      > Rickard.

      I'm not trying to put down flashboot (at all - I'm sure it works great for a lot of people), but have you ever considered just doing a full -stable install using release(8), PXE, etc? Customize your siteXX.tgz then run the filesystems in MFS which also sync MFS to flash on shutdown and/or via a crontab. Binary upgrades are just as easy either via installer or manually too...

      I guess I'm just curious on what the main differences are, other than having a 'cut down' version of OpenBSD? Personally, I prefer a full install (excluding games, etc.) on some of my WRAP and Soekris, with some of them set for XDM... >=)

      Comments
      1. By sthen (85.158.44.148) on

        > I guess I'm just curious on what the main differences are, other than having a 'cut down' version of OpenBSD?

        The whole kernel+OS combo is in a single file, so it's easy to keep multiple versions around - this means you can upload a whole new system in one go, without affecting any running software, carry out an upgrade, and rollback if there's a problem.

Credits

Copyright © - Daniel Hartmeier. All rights reserved. Articles and comments are copyright their respective authors, submission implies license to publish on this web site. Contents of the archive prior to as well as images and HTML templates were copied from the fabulous original deadly.org with Jose's and Jim's kind permission. This journal runs as CGI with httpd(8) on OpenBSD, the source code is BSD licensed. undeadly \Un*dead"ly\, a. Not subject to death; immortal. [Obs.]