FreeBSD 9.1 Release Announcement | Hacker News Reader

FreeBSD 9.1 Release Announcement

freebsd.org

134 pointsudp61 comments

For most simple servers/projects I use Debian/some sort of linux, so I'm pretty unknowledgeable but curious about the BSD side of things.

What are the advantages/differences of using bsd over any other linux distro?

Edit: For anyone interested, I actually Googled the question and found a pretty interesting article: http://www.over-yonder.net/~fullermd/rants/bsd4linux/01

ZFS (http://en.wikipedia.org/wiki/ZFS) is one advantage of FreeBSD.

Actual design and architecture is the main 'feature'.

That article is still fairly good of a read, with a couple things slightly out of date, which I will note here simply for pedantry or anyone interested.

BSD no longer uses CVS; FreeBSD specifically has moved to SVN for their primary revision control system, and CVS and related tools (such as csup) are deprecated and in some cases disabled. Along with that, linux kernel is now in git, not bitkeeper.

While source is still the "base" method of installing and updating a FreeBSD system, and it will certainly always be available; now, binary base system and add-on utilities are becoming more popular. Specifically, `freebsd-update` can be used to update the base system's binaries and source. `pkg-ng` adds `apt-get` like binary package management.

ZFS is definitely one of the major things that has FreeBSD on all of my systems. I even run ZFS on single drive laptops:

  pool: zroot
  state: ONLINE
  scan: none requested
  config:
	NAME        STATE     READ WRITE CKSUM
	zroot       ONLINE       0     0     0
	  ada0p3    ONLINE       0     0     0
  errors: No known data errors

What benefits ZFS brings on a single drive laptop? I was informed it is mainly designed for high traffic file servers with redundancy and so forth.

Snapshots

and checksums (bit rot prevention)

I'm amazed at how generally unreliable drives are at retrieving stored data with ZFS checksums in place to highlight the problem - it seems that on average every TB of storages has a problem with a block every half-year or so.

And every 4GB of memory or so if you have ECC turned on.

I run btrfs on my main arch installation. Have for ~6 months. If you get the fsck.btrfs tool (it is upstream) you eliminiate the only real showstopper with btrfs on Linux.

You get the online compression (meaning you use less space and read and write faster on mechanical storage because you have less to read / write at a time), you get the snapshots, you get subvolumes, checksums, etc.

I follow the btrfs development because I'd love to have at least the features of ZFS on Linux. However I would not dare (yet) to use it as a replacement for my current home fileserver (which owns ALL my data).

Does btrfs provide cifs export like ZFS ?

Edit:

from the btrfs wiki https://btrfs.wiki.kernel.org/index.php/FAQ#Is_btrfs_stable.... :

Is btrfs stable? Short answer: No, it's still considered experimental.

Even with a single drive you still benefit from many of its advantages!

Snapshotting.

Cloning.

Transparent deduplication.

Transparent compression.

Silent corruption detection.

Self-healing (when using copies=2 or higher, obviously the only way to provide redundancy on a single-drive setup).

NFSv4 ACLs.

  > BSD no longer uses CVS

Additional pedantry: This may be true of FreeBSD, but I believe both OpenBSD and NetBSD still use CVS. I think DragonFlyBSD uses git.

Another FreeBSD discussion a few weeks ago had a bunch of good answers to this question:

https://news.ycombinator.com/item?id=4895022

Jails. Recently I have configured a dedicated server to run nginx + php-fpm in jail: an entire jail consists of two static executable files, a few config files, and a few logs. There's nothing to pwn there, even if there's unpatched stack vulnerability somewhere. Jails are a very impressive security feature of FreeBSD.

Do you know how they compare to Linux Containers?

They are similar.

My trepidation with LXC would be from a documentation POV... Jails are very much known quantities, while LXC is newer, not known for great docs, and thus is probably easier to screw up.

See this HN thread for more discussion:

http://news.ycombinator.com/item?id=4015172

First, keep in mind that BSD is an operating system in the same way that Latin is a language: they each live on only in their descendants. Latin became French, Spanish, Italian, and a few others, and BSD became NetBSD, FreeBSD, OpenBSD, and Dragonfly BSD. There are different pros and cons to using a BSD operating system, depending on which one you choose. But the common themes are increased security and stability.

For a better answer, the websites of the 4 BSD projects give details on their focuses and strengths.

I like this analogy. OS X is therefore like English; an utterly alien world, colonized by a descendent of the original.

Actually, I'd say English claims the title of being the truest descendant of Latin. English not only directly uses descendants of Latin vocabulary, it also uses plenty of words from the other Latin descendants. It's kind of like a C++ diamond inheritance pattern.

English actually comes from West Germanic (not Latin). But you're right; there are lots of shared words between Latin and English, and this is mostly due to the historical relationship between England and France.

For anyone interested in high throughput/low latency networking, netmap ( http://info.iet.unipi.it/~luigi/netmap/ ) looks really exciting in combination with a userspace TCP stack (e.g. unetstack). Pity I have no applications that could use it, would love to experiment.

Not sure I see why you would want to use it with a userspace network stack, I am sure at that point kernel networking is going to be faster. Netmap seems more useful for bridging, logging and UDP applications to me...

Localizing packet processing significantly improves cache performance, in addition to avoiding copies and context switches. Receiving data from a socket almost always necessitates it being copied from a SKB to userspace memory, and prior to that from the driver's rx ring to a SKB. Grep for all references to 'cache' in http://lwn.net/Articles/169961/ which is Van Jacobsen's earlier work on a system of comparable design. Another (insanely more complex) implementation is http://www.openonload.org/

Ah thanks for that, hadn't realised that was a use case too, only started looking at this stuff the other day...

Wonder if it would be useful for a user land network stack that paralleled what it could b

Pretty nice feature list, for a point release. Make sure you check out the new stuff coming in FreeBSD 10: http://wiki.freebsd.org/FreeBSD10

More user-friendly page: http://wiki.freebsd.org/WhatsNew/FreeBSD10

Thanks. Can't wait for "SMP-friendly pf firewall". Yay!

It will be interesting to see whether variable symlinks can bring some Plan9ish filesystem elegance to *nix.

"Please note that precompiled third-party packages are not available for 9.1-RELEASE at the time of release. See the Availability section below for further details."

"With the exception of systems relating to the building and testing of packages, all FreeBSD.org infrastructure has now been brought back online. A full audit of the third party package build infrastructure code ("pointyhat") and package testing infrastructure ("redports") continues, and neither system will be brought back online until audits are complete."

I'm very happy to see the new ZFS enhancements imported from illumos, but I am very sad to see nothing about DTrace.

I use FreeBSD on my servers, most of them with ZFS. I wish there was some facility (like in Solaris) than as soon as a harddisk in a mirrorset fails, a spare is automatically swapped in. I will probably have to write that myself ;-)

Not necessarily :-) There's a ZFS fault management daemon up and coming in version 10. Look at the very bottom for the link to the source code in this page: http://wiki.freebsd.org/WhatsNew/FreeBSD10

Although my research in this case only consists of a quick google search, I can't seem to find any information on the Capsicum (sandboxing) project. I think I read somewhere that they were going to be publishing a userspace library / make the feature non-experimental etc in 9.1. However, I see no mention of it at all in the release notes. Does any one know what the state of that project is?

Where are good, popular and cheap options to run FreeBSD in a VPS? (From the three should I pick any two?)

Tilaa (http://tilaa.nl) is my go-to for a few reasons. It's not as popular/well known as it ought to be, at least in English.

There's also ARP Networks (http://www.arpnetworks.com), which seems to have a good reputation, and RootBSD (https://www.rootbsd.net), which seems expensive for what you get. I haven't personally tried either.

Tilaa looks excellent! Thanks for the pointer. Can you please describe your experience with them in some detail?

I don't have the energy to type up a big long thing. Been with them over a year no problems encountered in that time, network seems stable and performant, they do not appear to oversell. If you have specific questions I'll try to answer them.

I can give a huge thumbs up for ARPnet. They are based in downtown LA (and peer @ 1 Wilshire and I think they also are part of the Any2Net exchange from CoreSite), they also fully support IPv6 on VPC's as well as managed servers.

I've had both from them and have found support to be excellent. Honestly can't say enough good things about these guys.

I tried both CloudSigma and ElasticHosts and they seem to work fine with 8.x and 9.x.

(at least on 8.x I had to move kern.timecounter.hardware away from HPET)

http://www.daemonology.net/freebsd-on-ec2/

FreeBSD 9.1 Amazon ami has been released. I've been running on FreeBSD on EC2 for a while without issue.

VDS6 (http://vds6.net/) has some very cheap IPv6-only plans. They use BSD Jail for virtualization.

http://www.rackspace.com/cloud/public/servers/pricing/ $16.06 per month for 512 RAM (no traffic)

https://www.hetzner.de/

Which version of ZFS does 9.1 implement (could not find any info on that) ? Want to replace OpenIndiana b151 with FreeBSD 9.1 and reconnect my ZFS storage...

v28 - the last version released under the CDDL. You should be fine.

Thanks

Solaris Nevada b151 is on version 32 :(

Moved on to ZFS on Linux. FreeBSD does not support sharesmb.

Something wrong with Samba?

I really would like not having to choose between Gnome Shell and ZFS. :-(

How is device (ethernet, wireless, Intel GPU) support on laptops?

I'm running FreeBSD-CURRENT on a Thinkpad X220 and have no issues with any of the built-in peripherals. This laptop has an i5-2540M. The e1000 Ethernet is supported by the em(4) driver and I have a "Centrino Ultimate-N 6300" supported by iwn(4). The Intel GPU support was merged back to stable/9 in time for 9.1.

One of them wil be replacing my T61 then. Thanks for the report :)

Thanks, how about suspend/resume?

Right now suspend / resume is non-functional from a user's perspective: that is, suspend works, and I can resume, but the screen does not turn back on. I hope that this can be fixed for 9.2.

I'm running Arch w/ Gnome Shell on this Thinkpad T520.

I use ZFS on Arch (although not on this laptop, actually). I don't think you _need_ to choose?

How well does ZFS work under FUSE?

I have no idea. zfs on linux [1] is a different approach: It's a kernel module, out of the official tree (due to the license issues).

FUSE should work as well, I .. just didn't try. So ZFS for me is now a ~normal~ kernel filesystem, with some caveats about the external dependency and ignoring the big 'solaris layer' it needs to provide before zfs even starts.

1: http://zfsonlinux.org/

No kvm virtio drivers?

For 9.1 they are available as emulators/virtio-kmod; 9-STABLE has them in the tree (I used both recently).

Reply on news.ycombinator.com