Why can't I access MetaFilter?

September 15, 2006 7:33 AM

Is there a reason why I consistently cannot access metafilter (or any other metafilter site) 2-3 times per day? Granted, these usually are at high traffic times of day. Mornings when I get up, when I get home from work, etc. However, that's not always the case. At first, I thought the site was just down once or twice. Now, it happens every day, most of the times I try to access the site. Nothing has changed in my PC setup at any point, and I was getting regular access to the site without problems about a month ago. If I wait long enough when I get a "page not found" error, the site usually will appear. I apologize if this was already covered.
posted by theantikitty to Uptime at 7:33 AM (24 comments total)

Check the Status Page when you can't get on.
posted by dobbs at 7:36 AM on September 15, 2006

This Current Status wiki page is also updated more frequently than the main status page.
posted by cgg at 7:50 AM on September 15, 2006

The whole site has been down quite a lot in the last couple of weeks.
posted by cillit bang at 8:02 AM on September 15, 2006

wow! that's actually super helpful. Thanks guys!
posted by theantikitty at 8:15 AM on September 15, 2006

cgg's link [fixed].

What I'd like explained, what's the problem? Because from the perspective of somebody who doesn't know jack about building or running a website, it's hard to understand why MetaFilter is constantly stumbling while Slashdot and K5 and similar websites seem to be operating smoothly. Is there a reason that MetaFilter is more difficult to maintain?
posted by cribcage at 8:23 AM on September 15, 2006

dang it - how'd that happen? Thanks for catching that, cribcage.

Apparently, all the server issues started when the jobs page was launched (and the extra traffic that brought). Not sure why that's causing problems specifically, however...
posted by cgg at 8:31 AM on September 15, 2006

It's Colonel Sanders and his wee beady eyes. And the Queen. They're out to get us all.
posted by blue_beetle at 8:44 AM on September 15, 2006

all the server issues started when the jobs page was launched

*cough*

*choke*

*gasp*

that comment is the internet history equivalent of ABC's Path To 9/11 miniseries. metafilter has been falling over with stunning consistency since mid-2000. it has defied all attempts to stabilize, including hardware, money and geography. it has enjoyed only scattered periods of relative reliability. that mathowie has not forcibly torn every last hair from his pate is a testament to his infinite patience. given how regularly it fell over years ago with only a fraction of the current users, subsites and traffic, its recent track record qualifies as positively stellar.
posted by quonsar at 8:46 AM on September 15, 2006

What's the problem? Shit's breakin'. This is my interpretation of several sources.

The unofficial status wiki:

13 September 2006 23:54 UTC: the web server did an auto-update last night and rebooted and didn't come back up. Mathowie is working with the colo guys to reinstall the OS and get Apache running, but it's taking some time. He's hoping to get it up and running by this evening, California USA time.

A MeTa post from #1.

And an official status blog entry:

Bleah, ever since the jobs site went up, the server has been a bit unstable under mid-day high traffic. I've scoured the error logs and patched tons of little problems, and each time the server seems a little more stable but continues to periodically die.

I suspect that it might be a nifty interim solution to have a cron job just reboot servers every 8 hours or so.
posted by boo_radley at 8:47 AM on September 15, 2006

Man, these new users are adorable, what with their not knowing about the Good Ol' Days, when the site would go down regularly. Maybe the site should just start throwing Jrun errors every nth request, so we can kick it old school.
posted by mkultra at 8:51 AM on September 15, 2006

Every 12 hours, web services restart, so the site should be down for about 3 or 4 minutes at 6am/pm pacific every day, but yeah, there are still the ever-present stability problems.

The reason why slashdot and k5 stay up is that they have several programmers with real college degrees and everything that built real software before. Me, I just picked up a crappy language/platform when I wanted to graduate from photoshop and html.
posted by mathowie (staff) at 8:59 AM on September 15, 2006 [1 favorite]

All the server excuses are just a cover for the fact that Metafilter Hates You.
posted by ThePinkSuperhero at 9:13 AM on September 15, 2006 [1 favorite]

never underestimate the power of prayer
posted by matteo at 9:16 AM on September 15, 2006

worth noting: not only is the status page cgg/cribcage mention more frequently updated, it also tends to be reachable while status.metafilter.com can fall prey to the same conditions that are giving metafilter.com problems
posted by juv3nal at 9:45 AM on September 15, 2006

boo_radley writes "I suspect that it might be a nifty interim solution to have a cron job just reboot servers every 8 hours or so."

Assuming this even works we'd be trading 30-60 minutes of guaranteed downtime for the current situation. Plus rebooting is what took us down on the 12th.
posted by Mitheral at 9:53 AM on September 15, 2006

while status.metafilter.com can fall prey to the same conditions that are giving metafilter.com problems

status.metafilter.com is hosted on typepad, not this server, and has redundant DNS now. It should be completely independent of any problems on the existing server.
posted by mathowie (staff) at 10:33 AM on September 15, 2006

the site should be down for about 3 or 4 minutes at 6am/pm pacific.

6am pacific.... = 9am Eastern. Right when I usually try (and fail) to fire up metafilter. Everything makes sense now...
posted by inigo2 at 11:12 AM on September 15, 2006

I tend to prefer this to this, since the former tends to echo my existential frustration and anxiety, whereas the latter tends to almost evoke pity for the person trying to fix the problem. In any case, I usually just open both and press alt-tab ctrl-f5 alt-tab ctrl-f5 alt-tab ctrl-f5 until my fingers go numb.
posted by koeselitz at 11:47 AM on September 15, 2006

Mitheral: that's a possibility, sure. And it wasn't just reboots what done mefi in, it was automated patching that did it.
posted by boo_radley at 12:25 PM on September 15, 2006

quonsar has it. Hell, Metachat was practically created as a venue to complain about jrun errors, back in the dark days.
posted by muddgirl at 1:16 PM on September 15, 2006

ever since the jobs site went up, the server has been a bit unstable

I think I see a way out of this... post a job for someone to fix it! ;D
posted by scarabic at 7:57 PM on September 15, 2006

why MetaFilter is constantly stumbling while Slashdot and K5 and similar websites seem to be operating smoothly

One reason for this is that these sites simply have more hardware. I believe MetaFilter has just the one front-end server. Thus, when that machine falls over, the site goes down. If there was another front-end server, plus a load balancer, the front-end servers could be rebooted once an hour if need be with no downtime. Slashdot and K5 are big enough sites that I imagine they have far more than two front-end servers.

(In a multi-tier architecture, the front-end server is a Web server that serves requests directly to users. It talks to the back-end server, which is the database. Some systems also have a middle tier, which is typically a Web services layer that implements business logic. It serves essentially as an interface to the database so the Web front-end doesn't have to make direct SQL queries but rather can make requests like "give me this user's profile." The middle tier understands what a user and a profile is, so that the business logic can be separated from the presentation. MetaFilter seems to have the business logic all wrapped up in the front end, which is not that scalable, but completely understandable for a site that started as a hobby and has not been rearchitected since then.)

Even without the extra hardware, /. and K5 would probably be rather more reliable because they don't run on ColdFusion, but hardware can cover a multitude of software stumbles.
posted by kindall at 11:12 AM on September 16, 2006

Even without the extra hardware, /. and K5 would probably be rather more reliable because they don't run on ColdFusion, but hardware can cover a multitude of software stumbles.

Doubtful. Slashdot runs on PHP and K5 is (I think) Perl. You could argue that their codebases are better-written (my gut says Slashdot is, K5 isn't), but I think that hardware accounts for many orders of magnitude of difference here. And after that, OS. The only thing that runs "well" on Windows is ASP.NET, and even that's debatable.
posted by mkultra at 5:07 PM on September 16, 2006

Well, so far as I'm aware, Perl and PHP don't occasionally just stop working as ColdFusion does.
posted by kindall at 7:47 PM on September 20, 2006