Esoteric Curio - Comments

Steve Goodman: Systems Operations and Fishing.

nospam@example.com (Steve Goodman) — Wed, 05 Nov 2008 16:59:48 -0500

A very cursory google search lead to this page, which has several lessons of unknown quality, since I didn't walk through any of them all the way: http://philosophy.hku.hk/think/ I think the root of critical thinking is being to identify what is, and what is not, applicable to the problem at hand. Hence, critical thinking, as you noted, requires a lot of knowledge as well. You need a deep and accurate mental model of the system in play before you can think about it critically. Once you have that fleshed out model, you need to be able to develop a tree of possible causes from the problem symptoms, then traverse the tree looking for an option that fits the scenario, and build a more edges when you hit a leaf node. That's my graph theory model of critical thinking

Allspaw: Systems Operations and Fishing.

nospam@example.com (Allspaw) — Wed, 05 Nov 2008 11:08:00 -0500

I'm also interested in what you find in ways of teaching this sort of thing. Just like an older mechanic that can diagnose engine problems by simply listening to it run, great systems people are those who can drill down instantly into what's going on, or have an innate sense as to what will go wrong next. Running firedrills, I find, can be helpful, but doesn't really get people in the frame of mind for everyday practice. Maybe we need a Parris Island for ops.

Arnaud: Systems Operations and Fishing.

nospam@example.com (Arnaud) — Wed, 05 Nov 2008 03:26:36 -0500

> How do you cultivate and improve intuition? I guess there are books on the subject. This is what school does/should do to some extent, teach you subjects and how to use that information in your thinking process. Asking questions and leaving room for people to develop seem like sensible things to do as well (and I'm not saying you don't do that

Jeff Macdonald: Whirlwind

nospam@example.com (Jeff Macdonald) — Mon, 20 Oct 2008 20:48:29 -0400

Theo, love to hear your take on the cloud. Rest a week before jotting it down in your blog. Bake some cakes with the girls!

LineFeed: Choosing Solaris 10 over Linux

nospam@example.com (LineFeed) — Fri, 03 Oct 2008 19:19:54 -0400

I've encountered the remount readonly bug as well on debian under vmware. After what I've read, kernel 2.6.22 fixed this bug by means of a workaround, and this fix was included in patches for RHEL 4 and other distributions using older kernels. Possibly another one of Solaris strong points vs Linux is scalability on ~8 cores and up. But that's another topic.

Nicholas Lee: Zetaback. Respect.

nospam@example.com (Nicholas Lee) — Wed, 24 Sep 2008 20:56:22 -0400

If you are backing them up as files, that's quote clever. Probably without much work it would be simply to also store those files on Amazon S3.

Theo Schlossnagle: Zetaback. Respect.

nospam@example.com (Theo Schlossnagle) — Sat, 20 Sep 2008 22:37:16 -0400

Marcelo: The sends are stored (optionally compressed) as plain files in a directory for each remote host. They are not received on the remote end. This is actually more flexible. It means that the host running zetaback could be any OS (not just one supporting ZFS). The advantage of receiving them is making them accessible on the backup host (FS exploration). Right now we do incrementals of a fixed base, we'd have to start doing them only of the last snapshot. Suffice it to say that storing them as a file instead of as a ZFS filesystem was thoroughly deliberated decision.

Marcelo: Zetaback. Respect.

nospam@example.com (Marcelo) — Sat, 20 Sep 2008 18:38:32 -0400

Congratulations! Your backup solution stores the snapshots backedup in directories? Or the stream is received and imported on the Master node? I thinks your solutions seems really fine, but send is more reliable using a receive in the other hand.

Allspaw: Last second scaling hack

nospam@example.com (Allspaw) — Fri, 19 Sep 2008 18:12:24 -0400

Now that's a good story. It's quick thinking like that which separate the men from the boys. -j p.s. Sweet graphs.

Jignesh Shah: Zetaback. Respect.

nospam@example.com (Jignesh Shah) — Fri, 19 Sep 2008 13:25:07 -0400

Theo, Congratulations on "Zetaback" winning 3rd price in the OpenSolaris Community Innovation awards. Jignesh

Sam: Zetaback. Respect.

nospam@example.com (Sam) — Fri, 19 Sep 2008 10:08:26 -0400

When I click on the links in your article (when reading it via RSS in Google Reader) I get taken to some exit.php page. If I open the article in a new tab and click the links they work. Thought you'd like to know.

Ebrahim: Last second scaling hack

nospam@example.com (Ebrahim) — Wed, 17 Sep 2008 17:12:11 -0400

Nice. Maybe a cluster file system (such as GFS or Lustre) could help you more easily with your disk problem. Take a look at it. And OS-level clustering (e.g using OpenMosix) is another quick solution for CPU and memory.

Pedro Melo: Last second scaling hack

nospam@example.com (Pedro Melo) — Tue, 16 Sep 2008 05:44:19 -0400

Hi, nice. I've used something similar but with Perlbal. It has the built-in ability to try several URLs until one matches. Best regards,

Theo Schlossnagle: Choosing Solaris 10 over Linux

nospam@example.com (Theo Schlossnagle) — Fri, 05 Sep 2008 00:00:24 -0400

Actually, that's the entire point of blogging. It sucks 'cause I say so. The evidence is my documented experience.

Ishmael: Choosing Solaris 10 over Linux

nospam@example.com (Ishmael) — Thu, 04 Sep 2008 09:21:00 -0400

"That is plain crap. Ext3 is generally known for being one of the most unreliable filesystems." Sources please. You don't really think that just because you say it I'll believe it right?