Best practices in debugging an 'Unknown error'

Question

How can I debug an 'Unknown error' message that I get with every request in Craft? Logging says:

******************************************************************************************************
2015/01/16 14:03:36 [error] [exception.Craft\HttpException.503] exception 'Craft\HttpException' in /storage/www/imagecenter/acceptance/craft/app/etc/web/WebApp.php:873
Stack trace:
#0 /storage/www/imagecenter/acceptance/craft/app/etc/web/WebApp.php(218): Craft\WebApp->_processUpdateLogic()
#1 /storage/www/imagecenter/acceptance/craft/app/framework/base/CApplication.php(180): Craft\WebApp->processRequest()
#2 /storage/www/imagecenter/acceptance/craft/app/index.php(62): CApplication->run()
#3 /storage/www/imagecenter/acceptance/htdocs/index.php(16): require_once('/storage/www/im...')
#4 {main}
REQUEST_URI=/index.php?p=admin/actions/users/getAuthTimeout&dontExtendSession=1
HTTP_REFERER=http://acceptance.imagecenter.blinkdev.nl/index.php?p=admin/entries
---
2015/01/16 14:03:36 [error] [application]

******************************************************************************************************

Answer 1

Unknown error messages in the control panel almost always happen over an AJAX request and are usually 500 response code Internal Server Errors, so there isn't an convenient way to display a meaningful error message with a full stack trace to help debug.

But you've found the next best thing... the error logs in craft/storage/runtime/logs.

Usually if you see it on every page in the CP, it's from something like Craft trying to write a license file to the craft/config folder, but it can't. The error you shared is a 503 response, which means "Service Unavailable". You'd typically only see that if a request comes into the front-end of the site while Craft is in the middle of an update, or the front-end of the site is offline.

So I'm not 100% sure the error you posted is the one that's causing the "unknown error". I'd enable devMode, clear out all your log files, reproduce the problem and scour the logs again to see if anything else pops out.