Weblogs for Steve
For a ten minute period yesterday afternoon debian-administration.org was the 16th most looked up domain on the Bytemark nameservers. That's pretty impressive considering how infrequently new things are posted to the site.
In some ways things have stagnated because I've been busy but there is also a minor feeling that "everything is done".
For example I lookup the simple introduction to working with LVM at least once a fortnight for my own reference. There are a few other pieces I consistently point people at too, kinda my own personal "top ten" list of articles.
There are times when I think "I should write about Apache2", but then recall I already did that a couple of years ago. Granted things have changed, but they've not changed terribly much.
In conclusion I do feel that a lot of the things I'd like to document or writeup for myself have already been done here, and re-posting things would be both tempting and a little bit of a waste of time.
New topics are always interesting, and I guess a few small pieces on t-prot, etc, wouldn't go amiss. Not earth-shattering tools, but still moderately useful and helpful.
There have been sporadic comments here which have shown up twice - I think I've tracked down the source of this now and corrected the bug.
I'll keep an eye upon it for the future just in case it repeats.
]
Over the past few months this site has become a lot less reliable than I would wish. This unreliability has been caused by two things:
- Kernel issues.
- Site issues.
The kernel the host has been running, until recently, was the stock Lenny AMD64 kernel. This would frequently hang with messages of the form:
task master:26085 blocked for more than 120 seconds. echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
I appear to have solved these issues by upgrading to a locally compiled kernel of a more recent revision.
The second class of problems seem to be self-inflicted. The machine hosting this site is an Athlon64 X2 3800 with 2GB of RAM and 2x 200GB drives. Unfortunately it has recently started suffering from the dreaded OOM-killer.
I intend to spend a few hours over the next few days to reduce the memory used by the server - via a combination of reverse proxying, local caching, and apache/mod_perl tweaks.
Additionally I've begged the provider, my employer, to up the memory. So in the next week that will be increased to 4Gb.
A combination of code tweaks and increased memory should hopefully restore normal service.
Previously this site was accessible via IPv6 via the 6to4 tunnel - now it is using native IPv6 as provided by the hosting company (Bytemark).
Hopefully this should improve the IPv6 response times, by eliminating a couple of hops and round-trips.
I'd previously lowered the TTL on the appropriate DNS entries, so I hope this becomes live for everybody soon. If you're at the mercy of a bad DNS cache then I apologize for the temporary outage.
I've just been informed that the SSL certificate on this site had expired - so I've regenerated another one.
The certificate is still self-signed because nobody is throwing money at me, and I in turn don't want to throw money at verisign ;)
You can verify the fingerprint by following these instructions:
I've updated the webpages describing how to run the code behind this site yourself - the content hasn't completely been finished, but I'll work on it some more over the coming days.
The reason for the sudden update? We have yet another site using the code:
They're using a slightly older version of the code, but I think that might change in the near future. I'm not even sure if anybody noticed but I updated the way articles are linked on the site:
v.s http://www.debian-administration.org/articles/633 - the former is a lot more bookmarkable, although the latter will work forever too. Legacy links, love'em?
After a brief hiatus I've updated the code behind this site - there are probably only two user-visible changes:
- Weblog Editing : Tag Support
When you edit a weblog entry you've posted you can now edit/remove/update the tags which are applied to that entry.
- Link Changes
Although all past links to articles still work I've updated the default format. For example:
There are a few more changes which still need to be made to make polls and weblog entries have the same kind of link structure, but they won't take too long to finish I hope.
Behind the scenes some other things have been tidied up, and I hope I've avoided the submission of spam articles too!
I've also updated the planet to hide suspended users. D'oh.
This host has now been upgraded to Debian's Lenny release.
Please report any bugs you experience...
I think things are mostly OK, certainly the test suite output looks good, and the planet/etc is updating correctly.
The only issue I'm aware of is that "article read counts" are badly bogus, due to caching. Thats not a big deal and invisible to non-authors so I'm not going to lose sleep over it.
I've completed the first of my planned Lenny upgrades. It was mostly painless.
First of all I updated my sources.list file, then I ran:
apt-get update apt-get install apt apt-get dist-upgrade
There were no errors, no failures, and things seem good. Of course the machine I upgraded only runs one service (cfengine) so there wasn't too much likelihood of failure.
I'll leave it a week or so then upgrade this host.. Then on to my other boxes in rotation.
A couple of years ago when I wanted to setup this site. I looked around for the software to host it. I knew I wanted "a lightweight CMS a little like slashdot", but more than that I wasn't fussy.
Eventually I settled on a project I found listed on Freshmeat called YAWNS. This was suitable because it was simple, extensible, and coded in Perl.
Over a period of a few months I've updated, reorganized it, and generally made it mine. The original project disappeared from freshmeat, although it is still being used elsewhere, and things kept ticking over.
Recently the creator of the original codebase setup a new site:
The site? Written in perl of course. And using my version of his original software. All very circular.
Still for the first time I've been able to update the code documentation to list another known user.
(The code documentation is woefully incomplete, out of date, and that may well explain the reason why nobody else appears to use it publicly. Though I'm prepared to be surprised.