Older blog entries for lmb (starting at number 25)

I'm working on my slides for BrainShare Barcelona, which have to follow the Novell corporate style. So I have to do them in ... OpenOffice.

Do you know how much that sucks? You can't cut and paste text between presentations, because that will also copy small things like the bullets used etc. There certainly is a way around that, but it doesn't appear obvious to a CLI fool like myself. Galleries have to be imported one by one, which is quite helpful, because Novell has several of them in flat and 3D each; I love manual work, really. The color lists which I configured in OpenOffice for some reason don't show up in the OpenOffice graphics tool. 1.1.1 also kept crashing on me when I switched virtual desktops, but 1.1.4 seems to at least be stable. Yes, I know, maybe some of this is fixed in 2.0.0, but for these presentations we have to use 1.1.x. I'm not insulting OOo, I'm just venting, because I just wasted my Saturday on that. (Which is my own damn fault for having put it off for so long, but at least now OCFS2 is merged in openSUSE - it's amazing what you'll do instead of the thing you don't want to be doing...)

Typically, doing one of these talks involves cut & pasting the same magic point slides I've been using for the last couple of years (uhm, maybe I shouldn't admit to that in public) from various sources as appropriate for the target audience and then just going there, screw design, and talk during the presentation. That is the Linux way(tm).

If I really, really shouldn't have finished the presentation yesterday I might have been tempted to write a magicpoint to OOo converter.

The presentation, by the way, will be about all those Open Source tools for High Availability on SLES9 - Linux HA, drbd, OCFS2, LVS, Multipath IO et cetera, and a sneak preview about what's coming up with SLES10.

More discussion on heartbeat 2.0.1. Some friction on the project, which I guess is a sign for a lively one.

Post-2.0.1, which laid out the groundwork by providing notifications for cloned resources, I think we will tackle multi-state resources, which will allow us to fully and entirely model replicated resources, from databases to block devices. That will rock. We also want to look at GFS/OCFS2 integration, and I think 2.0.1 actually has all the features we need internally already.

To actually have a kernel to test that on, I'll probably look at merging both the GFS DLM and OCFS2 in a private patchset for the SUSE kernel after next week.

On the side, Marc, I also went for a ride yesterday with my beautiful, beloved and full suspension Steppenwolf Tycoon mountainbike. Managed to get lost in the woods (literally!), and after 50km, 3 hours (a forest isn't well suited for going fast) and 2500kcal later I found myself still 30km from home and decided that I was done for the day and hopped on a train back. And that given that I didn't ride much this year at all and had only planned on a light one or two hour ride... Sore? Me? Never. Oh well, good exercise. ;-)

And now, without further delay, on to my BrainShare Barcelona slides. Pity me. OpenOffice, not magicpoint. How evil is that. Thank god for the last minute, it is so productive!

(I really need to get it over with. As hinted at above, I plan on taking next week off and spend some days in Hamburg, also bury my nose in philosophy books and, hopefully, finding enough coherency in my thoughts so I can write a few pages again.)

heartbeat's feature-rich release 2.0.0 has caused a new spike on the mailing lists, or so it seems. People struggling to get to terms with High Availability Clustering and fail-over. I really wish a larger percentage of these would however bother with the available literature to acquaint themselves with the general concepts, and of course the literature on the project itself. Not to mention the number of bug reports which ignore the bug reporting hints, even though they are linked directly from the GettingStarted page.

Given that an estimated 80% of all failures of IT systems are caused by human error, admins would do very well to address their own education first when deploying a High Availability project of any kind, which if you misuse it will give you just more rope to hang yourself with, and in particular with a feature-rich tool like heartbeat 2.0, a shovel to dig your own hole and a sawed-off shotgun.

Of course, being employed by Novell, I approve of the opportunities for consulting and support implied by this, but as a participant in the Open Source projects, I find it slightly less than satisfactory.

So please, do yourself, your boss, and everyone else a big favour. Read first. Asking to understand is encouraged, but asking questions which indicate you didn't even bother to try and read the documentation are insulting. If you don't like our documentation, take a note it is a Wiki and you can help improve it ;-)

And if you come saying "This is mission-critical and my company depends on it" two days prior to having to go live and didn't bother with the fundamentals, well, tough luck.

One of the things I keep wondering about is how openSUSE will affect the enterprise and the Data Center. To me, focused on pretty much just that (and of course, High Availability like the very cool heartbeat 2 project), that is a highly interesting and relevant question. Many other bright and smart people (well, and some trolls) are looking at how this will make the home users mutually benefit from openSUSE, so let me spend some cycles on looking at these segments.

It is my hope that for one, with the "home servers" influence, there will be strong overlap in openSUSE for "low-end" features like software RAID and basic server functionality in general (webservers, fileservers etc). Virtualization comes to mind too, which is quite sexy for home servers. However, the high-end features remain separate - how many people are going to run clustering at home, unless they are nuts? Multipathing? 512 CPU SMP monsters? z-Series machines? Probably not too many, and even though features which were high-end yesterday are common today - but there's still a divider line.

(Though there's a chance that treating even a single server as a cluster of one will be an interesting path for the future; benefit from the same recovery and automation logic as larger clusters, just w/o the ability to failover - but if you so wish, just add another node or two. That might be interesting to pursue.)

So, one way might be that ISV/IHV engineers will participate in openSUSE to get those things they've always dreamed (or bitched, more likely) about fixed and improved. This could be a great opportunity for technically savvy shops, and short-circuit or at least ease a lot of the bureaucracy (which in fact all of the enterprise distributions have accumulated now, and the larger community-driven ones too - well, maybe that's not a good sign ;-) for certain low-key stuff which previously hasn't warranted to go through the big machinery. You got code which works, we luv you. In this way, this could be a great complementary channels for the established partner programs, which of course need to continue.

And yet of course, there's the hurdle that the enterprise distributions are in fact branched from the common tree at a given time, and then continue rather independently, and distinct from the community backed programs - at least that is what history shows so far. And then, with the tighter rules for enterprise processes, getting the community involved becomes harder again. But it might be a cool path forward, where the distributors become more of concentrators and "consulting shops" for the customers who don't want to hire their own hackers (or want their own IT staff backed by a team of excellent experts), but where other small shops and students might choose openSUSE Enterprise (my marketing is going to kill me for this one!) and "pay" for it by feeding back patches and enhancements - and still have access to the support organization if needed. In this way, this could be a better community distribution than without the evil big company involved.

So, I keep wondering. What could make a community-backed distribution more successful and attractive to those spaces? How will this affect it? This, in the end, is not just applicable to openSUSE but everything else, too.

And of course, these are tired, personal thoughts at the end of a long weekend by a mere geek, and if you think I have any chance of representing official Novell/SUSE position, I've got some spare km^2 of moon to sell at a premium.

I don't blog much, but maybe you all should be aware that after a couple of long months, the Linux HA project has put out heartbeat 2.0.0, a huge improvement over 1.x, capable of resource monitoing, a full-fledged dependency model and, recently added by yours truly, support for phase-of-the-moon affected rulesets.

Check it out here.

So the new hype is no longer Advogato, but Orkut. I am still wondering to figure out what the benefit for me as a user of those social-network mapping systems is.

If any of you knows, please do tell me and send me a mail.

And no, It's cool! is not a sufficient answer...

So, finally, the master level.

Just a status update: I still work on mostly the same stuff, for the same company, and it isn't boring yet. I hope that I'll get to do some really cool stuff in the next 12 months, the rewrite of large parts of the heartbeat resource manager - though I dread to call the current code that - being one of the more interesting challenges.

I also get to play with fun hardware these days, and you all should check out the http://www.umem.com/ NVRAM cards...

Otherwise, I'm bored. But easily amused by dry, cynical humor. Go read it all.

Mindless babbling in a diary. Good night.

Still not master. You all suck. Regards, lmb.

Duh. Still not master.

And time_10 still hasn't happened yet.

Oh well. Time for another heartbeat, just to show I am still around ;-)

After effectively over eight years, I have put up my personal website again on the weekend; much content still to drop in.

I think I took the last one offline in about 1997, when it had been embarrassing me for about 3 years already. Lets see whether I will do better this time.

And I was quite amazed; I had hoped that 8 years would have been enough time for web monkeys to come up with some useable toolkit, template system or whatever -- I was so wrong, or at least I haven't found it. Of course, everybody mourns in their documentation (which is usually wrong and incomplete) how much energy is wasted because every other webmaster writes their own, limitted scope toolkit, and how their solution will be the end-all on which everyone can standarize... But quite frankly, the crap has just grown more complex over time without actually improving. After having examined a few of them, I believe one has no choice but to do it again...

I would go as far as attesting that the fifty percent who apparently do not write their own but reuse someone elses' code are either damn lucky that someone else had exactly their needs, or are just too stupid to notice. And no, I will not guess which one is more likely.

So, despite my attempt to use a cool, high-tech toolkit to make my life so much easier, the website is again done how it always was: CVS, vim, Perl and SSI. I expect I will put some more in via mod_perl soonish, when I find time. Maybe, when I look again in another century, web technology will have happened.

Thanks for listening to my rant ;-)