OpenBSD Journal

how stable is OpenBSD?

Contributed by jose on from the uptime dept.

mirkuz asks: "Hi all, i use OpenBSD from years and i love this os, but in this years i always asked me how much stable the system is compared to the other system. looking in the internet i find nothing, please give to me help :) Thank..."

I've always found stability depends on what I was doing with a system, but overall haven't had any complaints about OpenBSD releases. Thoughts?

(Comments are closed)

Comments

  1. By Anonymous Coward () on

    IMHO, stability is on par with any other UNIX-like OS. Performance can be lacking, though. =

  2. By Anonymous Coward () on

    Go here and read this story . ;-)

    Comments

    1. By Anonymous Coward () on

      What was your point ?

      that OpenBSD does not scale as well as FreeBSD ?

      doesn't mean it is unstable.

    2. By Anonymous Coward () on

      Yes, I love my OBSD edge boxes, too, but I wish they ran a bit more efficiently. You can actually watch the box grown under a big Perl CGI startup on a lightly loaded Apache box. A Pentium 133 will actually hang the console for a few seconds while it shutdowns Apache.

      However, the issue at hand is "stability". Though perhaps not mutually-exclusive to "security" and "scalability", lack of stability (even on marginal hardware) does not seem to be an OpenBSD hallmark.

      And anyway, OBSD is fast enough for most of us. We got two OBSD boxes running nothing but Postfix and spam filtering between them, at a reasonably busy pace. Correction: *very* busy. Internally we use Exchange, with several co-located servers through-out the world. The OBSD servers handle *all* external mail, including incoming spam.

      Outoging spam I don't know about. I don't think we do that, but who knows?

      *shrug*

  3. By supabeast () on

    I've run 3.2 on my firewall for over a year now, without a single crash or lockup. Of course, it doesn't see much variety, as it's just filtering packets and running SSHD, BitchX, Lynx, and the occaisional nmap scan.

  4. By chris () chris@unixfu.net on http://unixfu.net

    openbsd has crashed on me just like anything else.

    updating software has been the real bane for me if it isnt just a fw or router box, as there is no easy way to upgrade packages. say like openssl.

    but though i know people that say it crashes more for them than others, and i know people that say it crashes less for them.

    i personally have seen it crash and lock up just like anything else.

  5. By Hardwark () on

    I've got 8 production boxes, firewalls and web/mail servers.

    They've been running for soon 3 years, with an upgrade and reboot every 6 months (some month after release).

    I've had a grand total of 2 crashes.
    One was hardware failure and the other was a bug in arp decode (it was on a lab net and the bug was found when a network tech was playing rough with a network analyzer/tester).
    (Bug found in /sys/netinet/if_ether.c, a fix was submitted and added to CVS).

    Now, I don't whip the crap out of these machines, but they do their stuff, year in and years out.

    // hdw, Exec Pope @ Kallisti

  6. By Anonymous Coward () on

    We have OpenBSD running an irc server. The machine crashes daily (such that a reboot is required). I was told OpenBSD was stable, but I don't see how it is.

    Comments

    1. By Anonymous Coward () on

      Okay, let me get this straight...

      You're running a service...

      Open to the public...

      That's known for being poked and prodded by those who are on it...

      And you're blaming the OS?

      I mean, if I'm off, let me know; otherwise I think your time would be a little better invested in finding a better IRC daemon.

      Comments

      1. By krh () on

        Well, he does say that a reboot is required, which sounds like it's the OS and not the IRC daemon.

        I can say that OpenBSD has been, in my experience, mostly stable. I was once trying to use it on an old 486, and there was a race condition in the Ethernet driver for the card I was using that caused crashes roughly once every two days. Getting it fixed took a very long time--I submitted multiple bug reports with as much documentation as I could, but unfortunately that Ethernet card was rather old and not very popular, so nobody bothered to do anything about it for months. Once it was fixed, the system was solid again.

        That experience aside, OpenBSD has not crashed regularly for me. If I were in a situation such as that described by the first poster, I'd try my best to figure out what's going on, and then submit a bug report.

        Comments

        1. By Gimlet () on

          A crash like that is either hardware or OS. From my experiences with OpenBSD over the last five years, I'd be checking out the hardware...

      2. By Anonymous Coward () on

        If the computer system crashes, the operating system is to blame. Always. That's my view. Nothing should bring down a modern system in software, apart from running things as root that is.

        Comments

        1. By Anonymous Coward () on

          ... because system software can always handle a hardware failure.

          It should be rare that a userland process crashes the entire system. However, things like device drivers and the like are in the kernel space (more-or-less). Failures here are nearly impossible to completely guard against. Panics and lockups due to hardware are endemic to any "modern" OS based on 30-yr old tech, like Unix.

        2. By Anonymous Coward () on

          Absolutely! Why, if the raid controller or hard drive fails the OS should always be able to recover. If you pull a DIM from the machine while it's running the OS should always be able to recover. If you pull a CPU from the machine while it's running it should be able to recover...

        3. By Anonymous Coward () on

          That's the most idiotic thing I've ever heard.

          Have you ever though about hardware? Electricity? Flaky ram is especially bad at causing intermittent crashes.

    2. By Teknoenie () teknoenie@site-fx.net on http://www.site-fx.net

      What would be nice to know is if the system really did crash; 1) did you submit a bug report to the OpenBSD crew, and 2) what were the specifics of the crash. It is quite possible the the IRC daemon was misbehaving and not the OS.

    3. By m () on

      i've run an ircd on openbsd since version 2.9,
      no problems so far and the only downtimes were
      for upgrades.

      i did have to tune the usual suspects in the
      kernel, running an ircd can be fairly taxing
      on GENERIC.

      Comments

      1. By Anonymous Coward () on

        For the people who might want to run a loaded server, what are the 'usual suspects in the kernel', please?

    4. By tedu () on

      would be nice if you filed a bug report or had told somebody about it before now.

    5. By Anonymous Coward () on

      What hardware is it running??

      I've got a dozen or so Squid proxy servers /w a custom PostgreSQL, Apache, PHP proxy log summary (by user, by site, by date). On Pentium II 400 /w 128mb RAM... and 400+ users...

      These servers have been in use for years (I just upgraded a v2.9 server to v3.4), without any problems.

    6. By Anonymous Coward () on

      Have you ruled out bad hardware? Are you running on some cheap PC or overclocked PC, or do you have a server with ECC RAM and well-designed cooling? Do you have components that might be using immature drivers?

    7. By jens () on http://jens.powerpuff.org

      at work run a OpenBSD3.0 server that runs httpd, ircd, icecast streams, mysql and some more. It has not crashed ones since we installed it!!!
      I think that guy whos machine crashes daily is either a sucessful troll, suffers from bad HW, need a braincheck or all_of_the_above

      regards /jens

  7. By Tony () aschlemm@comcast.net on mailto:aschlemm@comcast.net

    I've been running OpenBSD for print servers and firewall for the last couple of years and the systems I've setup have been quite stable. I do have one system that appears to have some problems with SMC EtherPower II NICs. When running the NIC in 100BaseTX, full-duplex mode the system would sometimes rollover and die and console simply shows a "stray interrupt 15" message. I tend to set the card back to 10BaseT and at that speed I have no stability issues in this system.

    I've tried swapping NICs with a different SMC card but the problem persists. I have SMC cards in other systems that don't exhibit this behavior so I don't know if it's an OpenBSD problem or I guess it's just the right combination of hardware causes this problem. If I've needed the system to be connected at 100BaseTX I've used a cheap Netgear FA311 NIC. Just in case it is a SMC problem though I tend to use the SMC cards on the WAN connection side of dual-home firewalls as the connection speed is 10BaseT on the broadband side and the cards are solid performers at that speed even in my one flakey system.

  8. By Ryvar () on

    Best rule of thumb I've seen for the mainstream OSes is:

    Win9x
    That's just a generality gathered from talking to hundreds of admins, swapping stories/advice/etc.

    There are good reasons for each of these, too - WinNT is (in theory) a server product and thus more stable than 9x, Linux is an open source product and many eyes makes for better debugging, FreeBSD is a cathedral-variety project and thus better on the stability front, OpenBSD isn't focused on performance and thus tops FreeBSD, and finally NetBSD is crossplatform so any bugs will turn be glaring on at least one of those platforms.

    I'm hesitant to place Solaris/HP-UX/AIX/OS/400 (especially Solaris) on that list because I've heard a *WIDE* range of tales as to where each belongs on it. Someone else with more experience with those four want to take a stab?

    Comments

    1. By Anonymous Coward () on

      Well our experience differs.

      We have/had OpenBSD installed on about 8 machines,
      on 4 it woruld work with no problems (inclusing the fw/gateway). However the webserver and the mailserver would crash once every month orso, the maximum uptime we had with OpenBSD was about 60 days.

      This unstability coupled with a forced upgrade every 6 months, made us change to Debian/Linux.
      On the same hardware (dell 1" rackmounts) we have
      uptimes well in excees over two years (database server). The linux machines haven't crashed (yet,
      two years of operations) and they feel more 'snappy', this coupled with the availability of more software (free AND commercial) and the 'wonders' of apt-get made us very happy.

      Comments

      1. By Anonymous () on


        This unstability coupled with a forced upgrade every 6 months, made us change to Debian/Linux.


        Wow, this statement floors me.

        I don't see anyone putting a gun to your head, to upgrade to the new releases. Guess what? Every OS has updates/patches. Just because Debian doesn't release a new dotted version every six months means it's more stable - haven't you applied patches since?

  9. By CMF () cfuhrman atta spamcop dotta net on mailto:cfuhrman atta spamcop dotta net

    I've run OpenBSD 3.2 on two Sun Ultra 1 workstations ... one a 143MHz model and another a 170MHz. I've had lockups during high network load on these machines which I understand has to do with a bug present in 3.2. I haven't tried it yet, but I believe this bug has been fixed in 3.4 (maybe 3.3).

    I also have OpenBSD 3.2 running on a P-90 w/32MG of RAM ... mainly a firewall/gateway box. Haven't had an *ounce* of problems with it which is pretty good for a relatively ancient piece of hardware.

    Comments

    1. By Tom Buskey () on

      Aren't there issues with getting UltraSparc documentation? OpenBSD on the older sparcs is solid.

      I'm always amazed at how well the various *ixen work on PC hardware. For example: sun has used between 5-8 different ethernet chipsets. How many chipsets are there for PC ethernet?

  10. By Simon Lok () simon@lok.net on mailto:simon@lok.net

    I have not had a big problem that
    doesn't show up during the installation
    or burn in. If it works when you get
    the machine up and test it for a week,
    it will tend to work for years (I'm not
    exaggerting, I literally mean, years).

    I perform updates when they're critical, and
    when they don't pertain I leave the machines
    be. I do all of my updating without access
    to the console. I have had machines with
    uptimes of over 365 days.

    The real problem is getting the boxes to
    go to begin with. At one point, I helped
    Theo by providing the hardware for the 82802
    RNG on the 82840 (the then brand spanking
    new 1.4 GHz P4s) to fly. I spent days
    trying to figure it out before I posted on
    misc@openbsd.org only to find out they
    didn't have access to this equipment before.
    Not their fault, just didn't realize stuff
    like that happened this way with OpenBSD.
    That was my first real schooling into the
    way things go with OpenBSD.

    Over the years, it has primarily been the hardware
    variation that have hurt... echoing my first
    experience related above. However, with
    enough effort, getting the first box to go
    has usually resulted in a "formula" that I
    can use to get other similar boxes to go. And
    like I said, once they get up and running, they're
    rock solid.

    Oh, BTW, everything I've been doing is with
    x86, so I'm talking about variability WITHIN
    a port, which you would not think would be
    as much as an issue. I have run all sorts
    of crazy stuff on the boxes, not just the
    built in stuff. Once again, if there's an issue
    with say, AMD, SQUID or some other service
    I just have to use, it shows up pretty
    quickly during installation or test.

    Comments

    1. By marklar_ () marklar_@hotmail.com on mailto:marklar_@hotmail.com

      >I have not had a big problem that
      >doesn't show up during the installation
      >or burn in. If it works when you get
      >the machine up and test it for a week,
      >it will tend to work for years (I'm not
      >exaggerting, I literally mean, years).
      Agreed!

      I've used x86, Alpha and SPARC ports since 2.7 and have found that things either "just work", or "just don't". This is unlike my experience with Linux and Windoze OSes where things seem to work for a while and then break, or don't work for some strange reason, but get fixed by the three R method (retry, reboot, reinstall).

  11. By Anthony () on

    I can't even bring Linux 2.6.0-test9 down.

  12. By Joe Price () on

    .. as well as that of others I know (because if there is problems I'm the first to know of course) I help administer about 15 OpenBSD machines for my company and 13 of those do pf/samba/dns/isakmpd/dhcpd/dhcpc/ntp.. Of those we have had only about 3 instances of unavailability in the last 2-3 years.. One was related to bad hardware, one was just a plain crash (dont remember), and one was more of an exploit, because the SSH didn't get upgraded and people were running an IRC proxy on it (not much trouble)..

    the other two run tomcat/jdk/apache/samba/pf/dhcpX/ntp/isakmpd/dns/spamd
    which is pretty much our web servers..

    I've had or have about 5 openbsd boxes at my home where I do various testing and things.. All my experience (since obsd 2.6) has been VERY GOOD. HIGHLY RECOMMEND OPENBSD.

    The biggest complaint with OpenBSD I have is the lack of a 1.4 JDK..

  13. By Anonymous Coward () on

    I have a similar question. Here it comes.

    I am going to be deploying a setup of about 5 (FreeBSD) web servers, 3 (Windows) media servers, and a (FreeBSD) database server. I'm planning on having these on a private subnet, with a firewall in front that will do NAT-style port redirection and load balancing. A simple, secure solution, assuming I can find the right firewall/loadbalancer.

    We expect the firewall will have to pass about 100mbit/s of traffic, and it needs to be up 100% of the time. If its down, we lose money, and lots of it!

    Under these circumstances, is OpenBSD a good choice? Anyone running a similar get-up? If so, what hardware is it on?

    A note, I have a lot of OpenBSD experience, but have not deployed it in such a mission-critical fashion. Just looking for people who have!

    Comments

    1. By Anonymous Coward () on

      Just another pornsite? 100% uptime is not possible. If you want to hunt for 99.9999% that is expensive. One single pc with any OS is not able to deliver this. You need at least two with no single point of failure in your network architecture and a reliable software. Or you buy two dedicated Loadbalancers like the ones from F5.

      Comments

      1. By Anonymous Coward () on

        I understand that 100% uptime isn't possible, but we do need a solution that will get close to it without breaking the bank. We can afford multiple machines, however last I knew OpenBSD didnt offer multiple-firewall combos.

        Comments

        1. By rankor_industries () on

          I haven't had time to play with this yet, but here is a deadly article that should help.

          http://www.deadly.org/article.php3?sid=20031018101733

          Comments

          1. By Anonymous Coward () on

            Yeah, that's what I was thinking too, but it sounds like that code is brand spanking new (not ready for a production environment).

            I'm trying to find a "tried and true" solution, something that won't cost me a plane ticket and a 3AM wakeup to fix :D

        2. By Dom () on

          If you want high uptime you need:
          Multiple BGP'd upstreams links (A and B), going out different sides of the building and never going along the same road/building/transit.
          Two switches, trunked together TWICE.
          A switch connects to A upstream router, likewise B switch/router.
          A pair of Cisco PIX, in failover, one on each switch. (or PIX blades, on in each 6500 switch)
          A pair of Cisco CSSs, one on each switch in failover mode.
          A web server on each switch.
          two of everything else, always one on each switch.

          All the 'A' stuff is on one UPS, the 'B' stuff on another. Both UPSs should be fed from (preferably) different supplies.

          I make sure that our two main datacentres are link this (better), as they did cost $10bn each :-)

    2. By Anonymous Coward () on

      Forget 100% uptime. No matter how much money you have to throw at the problem, it won't work. If you can live with "five nines", take a look at a hadware load balancer like the BigIP series made by F5.
      Remember: 99,999% uptime means only 5.3 minutes downtime a year. That means there is so much more inolved than the choice of your favorite OS. It helps to have a pair of diesel genererators on hand...

      Comments

      1. By Anonymous Coward () on

        It helps to have a pair of a pair of EVERYTHING around. Redundant everything including locations and network connectivity. Not to mention people on 24/7 to monitor everything to proactively find and fix problems before they happen.

        What the grandparent needs to do is determine the acceptable level of down time versus cost and go from there. Just saying 'no down time' is silly.

  14. By Anonymous Coward () on

    I have a similar question. Here it comes.

    I am going to be deploying a setup of about 5 (FreeBSD) web servers, 3 (Windows) media servers, and a (FreeBSD) database server. I'm planning on having these on a private subnet, with a firewall in front that will do NAT-style port redirection and load balancing. A simple, secure solution, assuming I can find the right firewall/loadbalancer.

    We expect the firewall will have to pass about 100mbit/s of traffic, and it needs to be up 100% of the time. If its down, we lose money, and lots of it!

    Under these circumstances, is OpenBSD a good choice? Anyone running a similar get-up? If so, what hardware is it on?

    A note, I have a lot of OpenBSD experience, but have not deployed it in such a mission-critical fashion. Just looking for people who have!

    Comments

    1. By Anonymous Coward () on

      What I'd like to know about this is how do you justify say OpenBSD (or even FreeBSD) over a commercial solution such as Cisco, Checkpoint, MS, etc. for security or DB's to the average corporate mentality or management for that matter?

      My experiences as a general network consultant are that management tend to prefer 'hardware' firewalls or other such commercial solutions rather than an open source, FREE, secure alternative - mostly because of commercial support and wider industry experienced people.

      So in this sense, I have a lot of hard times justifying OpenBSD over say Cisco when in fact for what most of them need, OpenBSD is a much better solution all around, from my experiences atleast. Don't get me wrong, I do believe in using the right tool for the job; be it MS, FreeBSD, OpenBSD, Linux, Cisco, or whatever. Many times I find that the options need to be balanced in either direction rather than just one or the other only.

      Did this make sense? If not, in essence what I'd like to know most is... How can you convince a consulting company, or management to let you design/implement OpenBSD in situations where it would be more cost effective and logical in a technical perspective?

      Any howto's on such a thing on convincing management or how to better market OpenBSD as an alternative solution to security than commercial products? ;)

      Thank you all in advance for those who understand and can help.

      PS: (To the poster above me here) I hope things go well for you however you get things done.

      Sincerely,
      Another *BSD advocate.

      Comments

      1. By Anonymous Coward () on

        How do you convince management to choose an opensource solution over a commercial solution? It's easy, convince them that they will have someone that they can BLAME when it brakes. The financial part is easy, but management loves to have someone on the line when things go wrong. If you serve yourself (or your consulting/contracting company) up as that person (liability and all), they will go for it. Of course, you might not want to take on that liability (things will go wrong and lawyers are sharks)....

        Comments

        1. By Anonymous Coward () on

          Very good point... ;)

      2. By Fred Flintstone () on

        What I'd like to know about this is how do you justify say OpenBSD (or even FreeBSD) over a commercial solution such as Cisco, Checkpoint, MS, etc. for security or DB's to the average corporate mentality or management for that matter? My experiences as a general network consultant are that management tend to prefer 'hardware' firewalls or other such commercial solutions rather than an open source, FREE, secure alternative - mostly because of commercial support and wider industry experienced people. I use both CheckPoint and OpenBSD depending on a number of variables. The main thing being price. I can buy a Nokia/Checkpoint solution with service and installation for $15k, or build up an old PC with OpenBSD for $0. But, understand that Nokia IPSO (the best OS for CheckPoint) is basically a BSD box running the CheckPoint software. So, without being rude, when you speak about "hardware firewalls", you're basically blowing sunshine out your arse. It's easier for me to deploy a CheckPoint box, and then centrally manage it, but with purchasing leading time, cost, and all those other factors, OpenBSD makes more sense in some environments (our VPNs are all based around OpenBSD rather than FW-1/VPN-1). Since I only use Nokia IP devices, that's all I have experience with, but these things are rock-solid. And they're basically x86 PCs. But, guess what? That old 400MHz P-II runs just as well.

        Comments

        1. By Anonymous Coward () on

          I agree with you on the "hardware" part, but most people from what I've noticed tend to think of a "hardware" firewall as a firewall that's dedicated to only that - not other normal I/O operations such as a PC or Server OS would do. In other words, a really cut down OS made to do only security and strictly nothing else. Some without mechanical parts, such as HD's.

          I agree that a PII-400 runs just as well, but I've had no luck convincing some management this in the past or that OpenBSD could be locked down even more than they think.

          I know the Nokia's use FreeBSD as the underlying OS, atleast last time I console'd into one and even use a HD as a normal PC would... Which is what it is.

          I also don't know why CheckPoint sells the software FW for Windows, Solaris and Linux but not FreeBSD if after all, it runs ontop of FreeBSD in the Nokia boxes? Perhaps because of marketing reasons?

  15. By Anonymous Coward () on

    I have had no trouble with OpenBSD thus far that wasn't caused by my own stupidity.

    That having been said, I have also not had any trouble with a NT 4.0 PDC, proxy server, and print server that I installed sometime in 1997, I think. It generally gets rebooted by local power failures, if anything. Go figure.

  16. By Gryp () gryp@dakin.be on http://gryp.dakin.be

    I've never had problems with OpenBSD except with my IDE drives. I guess the problem is my HighPoint controller, not my drives.

    Comments

    1. By Anonymous Coward () on

      I'll second this. Every box I've installed OpenBSD on (four so far, all cheap i386 leftovers) have had major problems with optimistic use of UDMA/DMA features. I finally figured out how to recompile with UDMA/DMA disabled (where appropriate) and everything runs just great. No speed daemon even with softdep turned on but stable, well thought out OS. Recommended.

  17. By kremlyn () on

    I've been running OpenBSD since 2.9 - its come along in leaps and bounds since then. The ipf drop for pf was the best thing that ever happened.

    I run OpenBSD at home only (unforutnately, I can't find a job that would have me administering a *BSD), however I have many machines doing many things with OpenBSD. I run:

    samba (for fileserving and domain logons)
    isakmpd (for home wifi)
    pf (duh!)
    httpd/mysql/php
    named (for local naming)
    central syslogd for all *NIX machines
    postfix/maildrop/courier-mta/sqwebmail

    I have never, ever had a machine crash on me. Not once. Admittedly, my hardware is tried and true.. all very stable PII based hardware with Intel NIC's. I guess as the story goes, the things OpenBSD does support, it supports VERY well.

    NOTE: I am moving to FreeBSD for non-internet-enabled services (such as samba) due to its SMP, ability to mount NTFS and easier upgradability. On the internet side though, there is NO substitute for quality.

    Cheers,
    kremlyn

    Comments

    1. By kremlyn () on

      I forgot to add a few things, I'm also using:

      ppp
      ddclient (for updating dyndns)
      dhcpd and dhcpc
      ntp
      ftpd/proftpd
      nfs

      All these work without a hitch too...

      Comments

      1. By Anonymous Coward () on

        There's support for NTFS in 3.4 now, it's been ported over...

        Comments

        1. By kremlyn () on

          Unfortunately, no SMP though.. but we're getting there slowly. Plus FreeBSD's I/O is much better than OpenBSD's..

    2. By Chris Humphries () chris@unixfu.net on http://unixfu.net

      maybe cause you havent really had them in production environments and such. boxes generally dont crash if they recieve very little traffic or requests.

      anything can sit there all day and do next to nothing, what matters is handling cruchtimes and heavy usage.

  18. By Anonymous Coward () on

    I tried installing KDE from the packages (pkg_add -v kdebase...tgz). However when I start kdm and log in, I get to an xterm with no window manager. Anyone knows what's going on? Did I forget to install something else?

    Comments

    1. By Anonymous Coward () on

      You forgot to install fluxbox... kidding.. ;-)

    2. By Anonymous Coward () on

      echo startkde > $HOME/.xsession

    3. By Anonymous Coward () on

      FVWM is nice and part of the base install for good reason. Configure XFree86 and type `startx' for a small, fast, and efficient window manager.
      WARNING! Don't let the "look" of the configuration scare you away, it can be changed. ;)

      Comments

      1. By Anonymous Coward () on

        The "quest for the perfect window manager" which starts from having FVWM as the first experience(which was a bad experience), to trying to justify the use of the larger(bloated, slow) window managers, to finally coming full circle and realizing FVWM feels like home.

        http://www.igs.net/~tril/fvwm/index.html

    4. By Chris () on

      $ cd ~
      $ cat .xinitrc
      startkde
      $ PATH=$PATH:/usr/X11R6/bin
      $ startkde

      That should work for you, as it just di for me.

    5. By Anonymous Coward () on

      ITYW:

      `echo "exec startkde" >> ~/.Xsession`

    6. By Anonymous Coward () on

      put a .xinitrc in your home and fill it with

      startkde

  19. By Juanjo () on

    The wi drivers sometimes hangs my PRISM 2.5 MINI-PCI based WLAN card. I post a bug repport and I applied a little patch that fixes the problem (I sent the patch too).

    Just reset via software the card when it has a timeaut and all works fine again.

    My OpenBSD experience is good... but I've not stressed the system and I've got by now the same reliability that with FreeBSD or Linux.

  20. By oblek () oblekNOSPAM@lug.stikom.edu on mailto:oblekNOSPAM@lug.stikom.edu

    we've been running OpenBSD for ~ 1 years, and running -current on production webservers, dns, proxy and mailservers, no complaint, serving about 100 people or so...just sendmail crashes under heavy load

    Comments

    1. By Anonymous Coward () on

      Why would you run -current in production???

  21. By Makaveli () makaveli@nwps.ws on http://-

    I used OpenBSD two months and had two system crashes during that time.
    First of all, I had problems with my Realtek NIC's, I had two of them on my server:

    *reboot after panic: uvm_mapent_alloc: out of static map entries, check MAX_KMAPENT (currently 1000)*

    Instead starting to tweak OBSD's kernel I decided to buy a new NIC, 3Com 3CR990-TX-97, which I thought was supported by OpenBSD (is listed on Supported platforms page)
    First I had this problem:

    txp0 at PCI0 dev 10 function 0 `3Com 3CR990-TX-97` rev 0x02: IRQ 10: fw not waiting for segment txp0: fw wait failed, section 0
    Oct 21 23:13:01 bsd /bsd: vendor `3Com`, unknown product 0x9901 (class network, subclass ethernet , rev 0x02) at pci0 dev 10 function 0 not configured

    I received a patch from one of OBSD's developers which helped me to get my card working...well, not for long. It worked 6 hours until system paniced and flooded this onto screen:

    "No free cmd descriptors"

    So I installed Gentoo Linux and everyhing has worked well so far. I still use OpenBSD on another server though and have had no problems.
    I love OpenBSD, I just wish it'd be easier to update e.g. from 3.2 to 3.4 and had better hardware support. And I think the OBSD community just need more beta-testers to test hardware and stuff like that.

    Comments

    1. By Anonymous Coward () on

      I suppose this is a common thread in this discussion. How can we discuss "stability" of an OS, and yet expect the kernel to be able to handle Realtek cards?

      Realtek doesn't even work on Windows , for crying out loud. It's not the ethernet part -- if the card works for you, it's because you have a robust PCI system. The bus-level stuff on the cards of totally broken.

      I'd use stronger language to describe how utterly garbage these cards are, but there are small children present (Theo, I'm looking in your direction).

      Oooh. A cheap jab at Theo from an AC. I'm such a l0s3r.

      Anyway, I don't know nothin' about the 3COM cards, but you were running an essentially untested driver patch.

      I've said it before, and I'll say it again: stability _starts_ at the hardware level -- all the way down to the metal. Build and test good drivers (which the teams seems to be able to do) that run on robust hardware, and OpenBSD is as stable as a rock.

      Look for stability on your old spare WinTel PC, and you may get it -- but only if you are lucky. Don't install it on my LAN, though.

    2. By Anonymous Coward () on

      I've used many realtek NICs, all different models on a multitude of systems and never had a problem once. You sure it's not your BIOS set to PnP or something in there?

      Comments

      1. By Benjamin Walkenhorst () krylon@gmx.net on mailto:krylon@gmx.net

        Both of my home machines (Athlon XP 2400+ for desktop, Athlon 700 for small server) have Realtek 8139-based nics.
        They don't perform exactly great, but so far they got detected correctly and worked, under a variety of operating systems: Linux 2.4 and 2.6, Windows 2000 Professional, FreeBSD 4.8, NetBSD 1.6 and 1.6.1, OpenBSD 3.2.
        Most of these systems I had installed for toying around (my primary system is Slackware 9.1 for desktop and NetBSD 1.6.1 on the server); so I can't really tell you how they behave in the long run under various OS'es, but the Linux-NetBSD setup works okay. Also, under Win2000 the card worked well.
        Performance is another issue, but I rarely put the cards under heavy load, so that's not a problem to me...

  22. By Anonymous Coward () on

    There are more things to being stable than just sustaining crashes. Crappy cheap PC boxen tend to crash with a couple of months intevals no matter what OS they run (although I've had most luck with Linux and FreeBSD, but I have no statistical significance). You have to find the good hardware you like and stick to that. "Server" style brand name systems generally works well, but costs more.

    However, when stumbling upon an old OpenBSD installation, maintaining it is a nightmare. Upgrading is hard if not impossible. Compared this with Debian and Gentoo which are rather easy to upgrade and sustain, even given the box may be a couple of revisions behind.

  23. By Anonymous Creepard () on

    I've never had an Open Box crash on me that was not the fault fo the hardware. I've been running Open on various firewalls since 2.8.

    My 2.8 box was a P90, with 48 Megs of ram, it was slowly dying the death of old hardware, at boot it somehow managed to use a random number generator to decide homw much RAM it had, (never more than 48Megs of coruse) but the sucker kept running long after that. the machine would not giv eup the ghost till I finally tried to upgrade it to 3.0, at that point it gave up and wound up in the recycling pile.

    Then there was the 200MMX that replaced it, which happened to have bad RAM, causing a segfault everytime you logged into the thing, yet it kept chugging along like nothing was wrong. It would run for 6 -9 months till it exhausted it's 128Megs of bad RAM before it would die. It usually got rebooted by hand before that. (power outages, moving the rack etc.)

    IMHO Open is a rather solid OS.

  24. By Anonymous Coward () on

    Well,

    I am experiencing some issues with 3.4 and 3.3. It happened twice, so I am not sure how to reproduce this. Anyway, the issue is that the termnial console will write in CAPS regardless of the CAPS LOCK position. Making it impossible to log in the machine. I can ssh and restart the machine though.

    As paranoic as we are, I am thinking if I have been root, and the criminal is trying to avoid usage of the terminal.

    Comments

    1. By Anonymous Coward () on

      try ctrl-alt-f2 and/or hit ctrl-D to bring up a new getty/login

      Comments

      1. By Anonymous Coward () on

        Cool, I've had this issue too and I've always done the ctrl-alt-f2.. but I didn't know about the ctrl-d there. I'll have to try that soon.

        PS: That should be added to the FAQ...

        Thanks!

    2. By djm () on

      Using tip? Try ^A

  25. By Joe Schmoe () on

    I've always had problems with the mouse in X. I'm using a Belkin OmniView SE KVM by the way. Problem happens like this:

    1. I get XFree86 config'ed correctly with the mouse.
    2. I switch to another machine (via the KVM) and do other things.
    3. I switch back to OpenBSD and the mouse goes crazy, as if it was config'ed with the wrong protocol.

    A reboot solves the problem, but that's not a really good solution.

    Comments

    1. By Anonymous Coward () on

      I've found that this happens with my KVM switch,
      a Trendnet TK400, if I use the keyboard shortcut to switch back to my OpenBSD box. When I use the button on the KVM switch itself, the mouse driver stays sane. Have you tried using the button on the switch?
      hope it works for you-

    2. By Anonymous Coward () on

      I have always had problems with low end KVMs like Belkins. I have found this to be the case with any x86 UNIX-y OS I have ran (Open, Free, Linux). It has to do with XFree's pooling of the PS/2 port coupled with the lack of a "keep alive" from the KVM. I have a Belkin at home. My OpenBSD boxes don't have X installed, so I don't hook up the mouse on those. My Linux box has a seperate mouse attached and it works fine with just the video and mouse. If you want seamless performance go look at enterprise-class KVMs on Ebay. Cybex, aspen, even an older Apex will work. I use an Apex Outlook at work and it handles any OS just fine.

      Comments

      1. By Anonymous Coward () on

        Doh! I meant to type POLLING not pooling!

  26. By c.roberts () on http://www.physicalplant.ttu.edu

    I've read all these story's about OpenBSD being stable/unstable. Here is my 2 cents worth of gasoline to add to the fire. I've been running OpenBSD since 2.6 all on the i386 platform. In all my time of administering unix servers, and I mean unix as in solaris, aix, (cough cough) sco and *BSD, I've been very proud of my OpenBSD boxes. So it's not the most robust and quickest os out there, but it has been the most trustworthy and stable OS that I've administered. At my current job, I'm plagued with MS everywhere. With MS comes frequent reboots, although some days are better than others due to solar flares and such. I wonder if some people's proclomations about OpenBSD's shortcomings are based upon their own mommoth lacking knowledge of it and what the goals are of OpenBSD? Most likely so. My guess as to what is the cause of the problems that people unfortunately experience with OpenBSD are mainly a lack of understanding of a excellent OS, lack of common sense, failure to read the FAQ's, laziness, and being acclimated to a less than mediocre OS with a plethora of fancy colors.

  28. By Anonymous Coward () on

    having used computers on a daily basis for the past 18 years, windows for 10, MacOS for 6 and open-source OSes for say four (Redhat, OpenBSD, Gentoo, FreeBSD) on low-end i386, in my experience OpenBSD has easily the best mainstream hardware support, documentation and base install.

    Of course, it is not really suited for building a rendering cluster. It is not the design goal. It provides a secure, stable base for system building.

    How well it performs depends on the function, the installer's knowhow and various unpredictable hardware interactions. However, that's the case for every OS, and I've never really expected my hardware to always work on the first go nor at full capacity. I've also never seen it do that for other people. If something doesn't work, consider changing it for something readily supported, funds permitting, costs a little more but so does frustration. You can also check before buying.

    Also, consider that the BIOS on i386 is an often overlooked source of problems.

    btw, I have an as/400 that ran for five years straight next to me which I still haven't switched on...

  29. By Chas () on http://rhadmin.org

    I've installed OpenBSD for several applications. The most aggrivating install was for a personal firewall on a Compaq Contura (486) laptop. I submitted a bug report, was told "hey good luck you're on your own." Was able to obtain a second, similar laptop, which solved the installation issues, but the system would not run with a second pcmcia (3com) network card installed. Moved the hard drive back to the original laptop, and presto, working, reliable system, current uptime of over 2 months. Upgrading to 3.4 is going to be scary.

  30. By sander () sander@skyberate.ent on http://www.skyberate.net

    We running openbsd in our network for shell server this due the amount of users who tryed to hack our servers in the past.

    0 times since openbsd.

    We have good uptimes with it.
    only mysql crashes at least every week.

  31. By Anonymous Coward () on

    Since OBSD 2.8, I have installed OBSD on more than 12 production computers. Most of these boxes has been very stable, some with over 500 days of UPTIME before I did take them down for an overhaul.

    However, the hardware is the determining factor in the success I got. For example, I seldom get any issue with the Intel Ethernet 10/100 adapters. However, some 3COM card has been more troublesome. The Broadcom Gigabit (build-in on the DELL P2650) work well but has strange reaction at time with some software such as ZEBRA.

    Finally, the latest hardware is not always very well supported. The DELL P2650 with a PERC 3/DI controler is generally working well but the controler will timeout at time.



Latest Articles

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.]