Written By:
lprent - Date published:
11:56 pm, November 2nd, 2009 - 13 comments
Categories: admin -
Tags:
Sorry to anyone who thought that The Standard hadn’t published anything today. The site had in fact been pretty active for a Monday. You got caught by a wee mistake by me*
I moved the site to a more accessible location on the server in preparation for updates in the free time between now and New Years. But I made a screwup in the permissions for the SuperCache directories and siteXML files*. It managed to feed yesterdays pages from the cache to some of our audience today.
Some browsers and proxy caches served up the old cached pages conforming to the site hints. Most requested new pages. I’ll be going back through the logs to figure out what exactly was happening – it should allow me to tweak more performance out of the hardware by shifting caching rules. Of course that will make it more obvious if I ever screw up the same way again.
But it is a lot more productive and transparent than letting it silently disenfranchise readers. After all I’m not a NACT elitist who finds that there are more favors to be gained if everything is kind of opaque and Key-like – like their versions of of the EFA and ETS (reminds me of the Muldoon era of gerrymandered electorates and SMP’s mostly) 😈
Then I would also be able to see the fault myself during testing. There was a lot of testing yesterday. But after it was reported by addicts in the afternoon, I tested on Ubuntu linux, iPhone and Windoze 2000/XP/Vista on linux virtualbox. I also tested on browsers, Firefox, Safari, Seamonkey, Opera, Konquerer, and even Internet Explorer versions 6-8 on appropriate platforms. I also tested on 3 different internet links for possible proxy errors. No errors. That was all that I had on me at the time, so eventually I went to the house of a reader to look at the damn error in the evening..
My only OS/X box is hooked up as a media PC to the TV at home. OS/X doesn’t run in a linux hosted virtualbox legally because of Apples crappy hardware restricted license. So following the prophet Murphy, that seems to be what was affected with both Firefox and Safari getting the problem. I’d be interested if other people had a Groundhog day that wasn’t on a mac…
All fixed now*
Lynn
* For the techies. I moved the code from a system location to one accessible to my personal login, and linked it back to the original location. I changed the user ownership of all files to my login and apache’s group on the server. The problem was that the existing relevant write and execute permissions for the cache were on the user not the group. They were octal 640 and 740 rather 660 and 770 respectively. So after I changed user ownership to myself wordpress didn’t have permission to write new cache files or to delete the existing ones.
.
The server will be getting hardware changes this evening starting at 10pm NZDT.
The site will be off line for some hours.
I thought I was getting cache material, but I was leaning towards blaming Telecom first 😉
But yeah, I got it on Iceweasel (Firefox clone) on Debian Linux.
Xtra or a local web proxy was my first thought as well. Xtras more than a decade instability of their DNS system does not exactly Inspire confidence. But it showed on the same mac laptop on orcon as well.
me too, I thought everyone went away for the weekend
(Opera on linux mint)
Happened to me but only briefly – I’m running Firefox. Seemed to work after the first couple of reloads.
Fix the Search thingy? 🙂
Blip: been trying to. I have now tested 6 search engines. The problem is the 130k and rapidly rising comments that either the engines don’t index or they fall over indexing or they chew too much time indexing.
I have hopes for a background search engine called sphinx but haven’t had time to finish testing.
Only winding you up, mate. These days I do the “text site:thestandard.org.nz” Google thingywotsit – seems to work alright so long as the bit I’m looking for more than, say, a week old.
yeah lynn i was wondering if a good interim kludge might be to just have Search as a google page with the ‘site: standard etc’ preloaded?
My experiments have indicated that we don’t get a particularly ordered mix. But I’ll try it out again… This search while complete, isn’t exactly the best ordering. Plus it repeats a hell of a lot as the name pops up in all of the tag search pages etc…
Still probably more usable than the current search though.
hmm, i see what you mean. on the other hand at least a google search can still easily handle a long string of terms to help refine the output.
anyway, just a thought. obviously you’ve already thought of that – surprise surprise.
I reviewed some search software in one of my previous jobs. Nutch and Sphinx came out on top. Both are stable and reliable, and run at the command line level so there’s no such think as crashing and falling over because of the size. They both index things nicely so searches from the site should be fast 🙂
I’ll have a look at Nutch.
The biggest hassle is that I have to get the pages down as a file for sphinx to scan. I stalled looking at the API for a in-memory way of passing the page. There is a plugin wrapper for Sphinx – but it isn’t particularly good.
However it does seem to be the only way I can maintain a good search engine without crippling the wordpress with the volumes of posts and comments whenever it has to do a reindex
Nutch is good – you can just give it the URL and tell it to index from the front page.