500 internal error

[kalliwoda]

I get 500 error just now (12:35 CET): Safari browser 4.04, Mac OS 10.4.11., opened o.k. 12:37

In case this might be helpful:
This was about 48 hrs ago (my last edit): Trying to save a change Safari "hung". After about 3 min. a "force quit" and restart I got the 500 error for about 2 min. Now if everybody else got the 500 error during this time...

[pml]

Mac OS X 10.5.8 on Intel, Mozilla Firefox 3.5.7: a few instances of the 500 internal error this morning, but the best error message of all was this, after committing a page preview:

ERROR
The requested URL could not be retrieved

While trying to retrieve the URL: http://localhost/index.php?

The following error was encountered:

* Zero Sized Reply

Squid did not receive any data for this request.

Your cache administrator is webmaster.
Generated Wed, 10 Feb 2010 22:04:05 GMT by mysquid (squid/2.6.STABLE18)

HTH! R, PML

[imslp]

At this point in time I have done pretty much everything I can think of. My limited amount of free time (if such a thing exists) also prohibits me from investigating this further. I'll have to ask for understanding.

The zero sized reply is another form of 500 error. If it doesn't go away after refreshing the page, tell me.

[daphnis]

I know you've already done what you can, so not to add any more of a headache, but I received the infamous 500 error for about 7 minutes on Chrome (4.0.249.89).

[ThatBadDog]

I'm getting this error on IE8 a lot.

I also getting this one about 50-75% of the time I click on a link (any link, even the home page):

ERROR: Requested URL could not be retrieved The following error was encountered: Access Denied. Access control configuration prevents your request from being allowed at this time. Please contact your service provider if you feel this is incorrect.

[imslp]

@ThatBadDog: Do you have some "internet accelerator" installed? Anyone else can test the IMSLP site on IE8? That error should not be triggered unless you or some other software changed the max connections limit in IE8.

@Daphnis: I have a few new insights as to what the problem could be. I don't have time to pursue this matter presently, but will do so soon.

[homerdundas]

More problems... more 500 errors and now

Unable to connect
Firefox can't establish a connection to the server at imslp.org.

Have the enemies of the public domain finally scored a direct hit?
(this trashed my upload, yet again )

[Notenschreiber]

I have received this error information too, not the first time, but now longer than ever before.

[imslp]

That was entirely my fault. I felt that I had my finger on the reason for 500 errors, so I took the server down to fix it. Unfortunately the fix took longer than anticipated, hence the downtime.

If you get further 500 errors after this, tell me.

[Carolus]

Although the wiki has been behaving OK for me lately, I just noticed this oddity in the traffic report: http://imslp.org/stats/usage_201002.html Has the traffic really fallen off that much? Seems strange as the number of downloads has continued to rise. Also, I just noticed that the link on the "Sitemap" page to the stats does not appear to function.

[imslp]

Actually, no the traffic hasn't fallen at all. The traffic "dip" you see was because I made the site work with caching software (called "Squid"). In essence, approximately 80% of hits are never processed by the webserver, but by the cache. That is why you see the traffic "dip." The cache has essentially turned one server into five servers. Well, that's a slight exaggeration, but not by much. Thanks to the caching software, our current server can probably last another year unless the traffic explodes for some reason.

The reason the stat link doesn't work on the sitemap is because there is no trailing slash (/ at end). This should be fixed by someone...

[Choralia]

At CPDL we had a similar problem in November, when the traffic raised to some 500,000 - 600,000 page impressions per day. Our inexpensive shared hosting account showed some 4% "500" errors per day, with peaks of some 10% during rush traffic hours.

Having no budget for a dedicated server, and being quite difficult (or probably impossible) to add Squid to a shared hosting solution, I hacked the MediaWiki code so that registered users who are willing to add scores, edit pages, etc., are directed to the "contributor website", running on the original shared hosting account, while all traffic from non-registered users is distributed towards any number of read-only "visitor websites" (we currently have two), running on additional shared hosting accounts. The cost of multiple shared hosting accounts is far lower than the cost of a dedicated server (or of a powerful VPS). As most of the traffic is from non-registered users, this solution demonstrated to be very effective, and it's fully scaleable to accommodate traffic growth. Visitor websites are automatically aligned to the contributor website on a daily basis, so they implicitly provide back-up and site diversity/disaster recovery features.

IMSLP certainly needs much more powerful resources, anyway I wanted to share this information if of any interest for IMSLP.

Max

[imslp]

It seems that IMSLP still has 500 server error issues. The 500 server issues CPDL experienced was probably because of resource limiting (the host limited the resource usage for each account). In IMSLP's case we have two dedicated servers, the main one being a Quad-core Phenom. But I think the problem is not the CPU (currently only 50% loaded at the busiest with Squid), but that the hard drive just simply cannot keep up. So we have downtimes of ~4-5 minutes per day, which is still acceptable.

I may turn off FTP access for a while in the near future to see if that is the cause.

[Carolus]

It's 10:10 PM Friday CST. The wiki appears to be down, as Safari suddenly gives the message: "Cannot connect to the server: imslp.org". Same message when attempting to go directly to the stats page ( www.imslp.org/stats/ ). Just did a traceroute and got this: "traceroute: unknown host http://imslp.org/wiki"

[daphnis]

Same here, although it seems to be back up now.