Error 503 Service Unavailable on many pages

Getting this again this morning. some other pages are a bit slow to load.

Some naughty queries put stress on the DB server. We've implemented a new memcached bin to allow for better (and longer) caching.

Combination of load balancer "issue"/feature and varnish timeout. This should be fixed, but please report if it happens again.

Thanks.

How long did it last, what page, did a force refresh help?

Also, next time anyone sees this can you give me fairly exact times and your IP. I am seeing actual 503 errors coming from apache, so this maybe totally unrelated to varnish.

Hass-

About how long did it take before the 503 came up?

Thanks for all this info btw, very helpful.

We have edited the varnish config a bit to ensure that 503s are not at all cached and to up the timeout. Let me know if this continues to happen.

I believe I'm going to close this. I'm fairly convinced this has nothing to do with varnish, but is just a symptom of times when d.o is overloaded or a query hits a timeout. Thanks for all the reports. This should be considerably improved by the infrastructure improvements we are currently working on.

This started out as an issue with varnish config, but now these absolutely line up with apache 503s. Thanks again.

-N

I am still this error on all Drupal forms.

Subscribing...

I've been getting 503, WSOD, and partial pages (text only - no css or jscript) continuously for the last week. This has been especially bad over the last hour. (75% of pages) I've hit F5 about 20 times in 30min and the form that I am trying to submit still is not saving.

Its been almost a week now I've been getting these:

Error 503 Service Unavailable

Service Unavailable
Guru Meditation:

XID: ????????
Varnish cache server

I use to get this error creating new comments on issues. But in the last 2 days ocurred 3 times doing a normal navigation.

(I guess this vague comment could be taken as a +1) :-D

What was fixed? This seems to be a consistent error on poor connections. I experienced this on a regular basis when on the road in South America and often get it when the wireless (mobile) internet connection in Australia is poor. I have seen no change in this over the last year and a half.

I can accept that the WSOD / 0 Byte responses could be, but a 503 != local access error.

I'm fairly convinced this has nothing to do with varnish, but is just a symptom of times when d.o is overloaded or a query hits a timeout

On this 2 year old response, and minimal noise about this, (developers can hit F5 in their sleep), marking as a won't fix as I do not believe that this is fixed, but I haven't the time to investigate further atm and will be getting a real connection within weeks :)

Hello,

I have solved this issue. After watching the varnish log file you can see this issue when the Drupal cache is hit.

16 FetchError c Junk after gzip data

The issue was not .htaccess enabled gzip compress but he Drupal gzip compression.
To solve this issue I have done these steps.

Open settings.php and add these lines

$conf['css_gzip_compression'] = FALSE;
$conf['js_gzip_compression'] = FALSE;
$conf['page_compression'] = FALSE;

Then login to domain.com/admin and Flush all caches. After doing this and logging out the site is no longer having any issues and is working as expected.

Hopefully this works for you.

In my case the Cache-Tags header had gotten too big for Varnish. The default is 8K (or 8192 characters, I believe) and the actual header had 8491 characters in it, after only about 700 nodes. Adding the parameter http_resp_hdr_len=16k to /etc/systemd/system/varnish.service.d/customexec.conf and restarting Varnish fixed the problem.

I solved this issue, I had the same problem on all /admin/* pages on a Drupal7 site.
In my custom theme, I had a folder called "node_modules" that I forgot to mention in my .gitignore file. On my local environment, this wasn't a problem, but on the production server, I got a 503 error.
After removing node_modules from the repo, everything was fine again.