Comment FreeBSD? (Score 4, Informative) 155
If you want Dtrace and ZFS, just go with FreeBSD. You get pf and jails thrown in for the effort.
If you want Dtrace and ZFS, just go with FreeBSD. You get pf and jails thrown in for the effort.
Pick your country and protocol. Works with the BBC, Hulu, Vimeo, Pandora, Netflix, probably just about everything else. Of course, you have to pick a US proxy for Hulu, a British proxy for the BBC, etc.
Many FOSS projects have their primary development done on xBSD, so that fear may be overstated:
I could not agree more. I have used many of the GEOM features, they all work perfectly. The partitioning scheme, in particular GPART and GLABEL, make working with USB and other hot-swappable devices much easier and less error prone.
The FreeBSD motto is "the power to serve", linux laptop users should keep that in mind.
I had not seen that article yet, thanks. However, it still misses the point. ZFS is *not* intended for 4GB workstations with single hard drives!
I'd like to see them benchmark a ZFS setup with 3 or 4 RAIDz pools concatenated into a 10 to 20 TB partition spanning 10 to 15 physical drives, with an MLC L2ARC. Throw 48GB of ram and an 8 core processor at it as well. Next, properly tune the InnoDB and logging partitions. Then, benchmark MySQL. Finally, try a large update/join spanning many tables, using foreign keys and stored procedures-- and:
-- pull the power plug, see if there is any data corruption
-- do a ZFS snapshot, live, see if the benchmark notices the difference
-- add more RAIDz pools to the live file system
People choose ZFS for the reasons in this article:
http://www.zdnet.com/blog/storage/zfs-data-integrity-tested/811
Sorry to hype on ZFS so much, but that is where my expertise is. Their benchmark was testing a scenario that nobody would be encouraged to ever actually use in the real world. It makes me seriously doubt the usefulness of their other benchmarks (like HURD) when they make such an obvious (to me) mistake as to omit these important details.
Phoronix has a history of questionable choices for their benchmark setups. Hardware, versions, and tuning are... cleverly chosen, almost as if there was a preconceived agenda with inevitable results. Not that there is one-- just like it seems like there is. And so colorfully presented! I remember when they tested ZFS on an i386 version of FreeBSD on a 1G laptop! Others have also noticed this Phoronix phenomenon:
http://forums.freebsd.org/archive/index.php/t-16396.html
http://www.kev009.com/wp/2008/12/phoronix-benchmarking-statistically-significant-and-other-performance-concerns/
The whole point of Hurd, at least right now, is tangential to benchmarks. Nothing wrong with testing, of course, but I think the results should not be used for any long term planning. Nobody is planning on launching a business running on Hurd servers... yet.
Stupid... but cool as hell. There is such a fine line between stupid and clever.
PF + AltQ, a ZFS raidz array, and booting from a CF card. Excellent job, kudos to the FreeBSD team!
I find it hard to believe that the folks whining (I'm sorry, "bitching") about sudo usage are sysadmins on servers, and certainly not servers that are depended on by others. This policy is a good idea on any system that you can access remotely (thus making it a "server"). Running an internet connected server like a five year old is selfish and it should not be a surprise that it is discouraged.
Presumably when doing system operations, you will do as little as root as possible. Therefore sudo is not much of an inconvenience. Yes, you could prepend a destructive command with sudo, but you would have to be twice as stupid.
If remote root logins are disabled, then you cannot (remotely) guess the root password.
I read your post as:
"I am so good, and so careful, I would never, ever make a mistake as root."
Good luck to you on production servers, and may your employer and clients have mercy on your soul.
Look, admit it: running commands as root is a convenience for you, and you are willing to make the obvious tradeoff in stability and security. But don't imply that others are as gifted as you are in avoiding simple mistakes that are catastrophic as root.
This touches on another point, that is being "root" at any time other than sysinstall. FreeBSD has never (by default) allowed root logins via SSH, and I will always contend that is a "good thing". If you access a system via SSH, it is a server. If you are on a shell session on a server, you should NEVER be root-- that's what sudo is for.
If you whine about this, you are indeed a poor sysadmin. It reminds me of my friend who habitually texts while driving. "But I have never been in an accident," he says. How selfish, putting his convenience above the safety of those around him.
I found that when a client of mine connected via SSH to a well connected server (Equinix/Ashburn), they could use the SOCKS setting in Firefox (connecting to localhost since that's what their SSH client listened to) to tunnel all of their traffic with no problem. Note: this was a Mac, up to date as of last year when we tried this.
Sure enough, one day the tunneling stopped working! We changed the port used by SSH to 443, and it worked just fine after that.
What is the responsiveness of the system under load? Openssl speed? bonnie++?
All seems condemned in the long run to approximate a state akin to Gaussian noise. -- James Martin
