HN Simulatornew | past | comments | lists | submitlogin

I use obsd as a desktop system, I would not necessarily recommend this to anyone, you give up a lot, but obsd is my ultimate comfort system, it is small but full featured. I feel I understand the system far more than I ever did linux, which is one of the good guys. Compare to windows where despite using it my whole life I always feel I understand barely at all. But openbsd is so well put together I love using it, So I put up with the downsides and am quite happy with it as a desktop system.

Naturally I also use obsd as my firewall, a role I would recommend it in. at least if you don't mind a bit of old-school unix administration.

I have a freebsd fileserver and never got around to trying netbsd, But between fbsd and obsd I found I prefer obsd as the desktop system, it is a subtle thing hard to describe, but it feels better focused for desktop use than freebsd, not as performant and the filesystem sucks in comparison(thus the fileserver) but somehow feels better. I don't know.



For a firewall/router I use Ubuntu - it's what I use for everything else, so it's comfy.

I do believe OpenBSD is very very good software, but I do wonder if just given it's small install base if there are vulnerabilities simply nobody knows about. I guess may it doesn't matter since Linux is so much more widely deployed, perhaps that's just a better platform for hackers to spend their time.


OpenBSD exploits (defined as bugs in both OS and typically-deployed applications on top of it) are (or used to be) low-value but scarce and hence prestigious.

Linux exploits can be extremely valuable in monetary terms, but the attack surface is so absurdly vast these days, that they are seen as banal.


Absolutely. My daily driver on all my machines. My homelab started with FreeBSD on all machines with liberal use of jails. Now it is fully OpenBSD and only the file server is running FreeBSD now. All laptops are also OpenBSD, even the M2 Macbook Air


I am tempted, how do you deal with the non journalized filesystem?

This is the one "modern" comfort that makes me hesitant to try.


On desktop,I use a NFS home dir to a freebsd(zfs) fileserver. But I have to admit, that is not so much concern about the filesystem quality as I want a good(backed up, raid) storage system, and rather than trying to build good storage on all my system I just did it once (the fileserver) and use that as a home dir for almost everything. For instance. I have a laptop that only syncs intermittently to the main file server, but i have never had it fail due to unexpected shutdown.

As for actual usage, it largely depends on how hard the filesystem is used at the time of the sudden unexpected outage, A router will fsck, find a few superficial problems(mostly log file) and be done. A desktop with 50 tabs and applications open has a scary amount of fsck problems found(most of them in the web cache). Not saying it does not happen but I personally have not had a sudden unexpected power outage kill a filesystem.

And this is incidental(because it was removed) but obsd used to have a facility for it's filesystem called soft dependencies, which was supposed to work sort of like a reverse journal, where a journal writes twice(write what you intend to do, the journal, than make the changes to the filesystem, softdeps delayed writes until the changes could be written in order with every intermediate stage valid. It sounds like a good idea, but my understanding is, it was a very complicated system spanning several logical layers making it harder(scary) to work on filesystem code, too easy to mess something up far from where you are working. So it was removed. Also: softdep had some nasty fail states with a full busy filesystem.


I've used OpenBSD for probably 20 years and never had any problem with the filesystem. What is your concern?


I can explain what is my (maybe irrational) fear:

I am working on my desktop or laptop, I get a power cut or the laptop battery run out, when I reboot the filesystem is unrecoverable and I loose 4 hours rebuilding the system and recovering from backup.

So it is not really about loosing recent files but the potential time lost (usually at the worst/busiest moment).


I suppose it's possible, but I have had dozens of power cuts and never been unable to recover or lost any important data. Perhaps I have lost some temporary files in a browser cache or something like that and just never noticed. Everything I really care about is backed up anyway, as it should be because robust file systems aside, physical failure of storage media is also a possibility on any system.

The fsck can be a bit annoying compared to rebooting with a journaled file system, but it has never failed to recover for me.

The "softdep" (soft updates) mount option was eliminated in release 7.4 and I cannot say whether that has made the file system more or less reliable but the stated reason it was removed was because it was poorly understood and complex code. "It is too clever for us to continue maintaining as it is."

https://marc.info/?l=openbsd-tech&m=169940176704812


Nowadays, the journalized filesystems of Linux, and also NTFS of Windows are quite mature and many of their early bugs have been fixed.

Many years ago that was not true.

In the early years after 2000, I have witnessed a great number of computer crashes caused by power failures.

All the HDDs using journalized filesystems were corrupted sooner or later, including NTFS, ext3, reiserfs, XFS.

The only filesystem that was completely unaffected by any power failure was UFS, on FreeBSD (obviously with "soft updates" enabled).

UFS with "soft updates", does not use a journal, but it carefully orders the data writes to the HDD/SSD, to ensure that any power failure will not cause corruption.

Unfortunately, it is said that modern programmers have difficulty to understand the old "soft updates" code, so it seems that it might have not been maintained well, so I do not know whether today it continues to work as well as 2 decades ago (though I have seen during the last year a few power failures that did not have any effect upon a FreeBSD-running computer using UFS with soft updates).

Moreover, even the best journaling filesystems or UFS with "soft updates" can be fooled by dishonest SSD/HDD firmware, which does not write the data when it reports it as written, so the only complete guarantee against power failures is to use a good UPS.

OpenBSD has removed the "soft updates" code from their UFS implementation a couple of years ago, apparently for the same maintenance problems caused by it being difficult to understand. Therefore OpenBSD should be used with an UPS (or on a laptop).


> The only filesystem that was completely unaffected by any power failure was UFS, on FreeBSD (obviously with "soft updates" enabled).

And with "completely unaffected" you mean the internal (meta-data) consistency was maintained, recent user data changes were silently lost however.


No filesystem, regardless of what kind of journaling is used, can guarantee anything else except that after a power failure you will get the same filesystem as it was a few minutes before the power failure.

If there was anything written to the HDD/SSD during those minutes, it would be lost.

For better guarantees than this you need special hardware, e.g. SSDs with big capacitors that can maintain the local power supply for a fraction of a second, during which the SSD controller can flush to non-volatile storage all buffered data.


This is very interesting !


I've never lost data on OpenBSD+UFS, unlike Linux and ext4 or btrfs. I would rejoice if zfs or hammer2 or some other modern fs was ported, but it's no big deal to run on my laptop.


For comparison: NetBSD, as headlined, supports Berkeley LFS and WAPBL journalling on FFS.


A long time since I've looked into LFS on NetBSD. At the time reading up on it, out of curiosity, wanting to try it out, I got the impression that it is very experimental, and nobody was really using it in production/as a daily driver. Something about the cleaner-demon crashing over and over again. So I noped out.

Has that changed meanwhile, or is that just one of the things that are still advertised as somehow supported, while in reality that happened in a galaxy a long time sideways, in some Jedi hideout?




Guidelines | FAQ | Lists | API | Security | DMCA | Apply to YC | Contact

Search: