could not build the referers_hash
jdorfman at netdna.com
Fri Mar 30 16:08:54 UTC 2012
Here are the current values:
The number of vhosts we have on that cluster is 23,252 (not all active,
need to clean that up.)
Are those values sufficient for the amount of vhosts?
FYI We have 32gb of ram if that makes any difference.
Justin Dorfman <http://www.twitter.com/jdorfman>
On Fri, Mar 30, 2012 at 8:17 AM, Cliff Wells <cliff at develix.com> wrote:
> On Thu, 2012-03-29 at 19:04 -0700, Justin Dorfman wrote:
> > Hey Cliff,
> > I did, same issue. The way we ended up "fixing" it was deleting all
> > of the vhosts and syncing them from another PoP.
> How large did you make the value? With a large number of vhosts the
> value might need to be significantly higher. Also, I'm not sure of the
> context for this value, it might need to be at the http level.
> There's also a related value, referers_hash_bucket_size that might need
> to be increased.
> The only information I found was from this thread (translated from
> Russian via google):
> 2011/7/19 Igor Sysoev <igor at xxxxxxxxx>:
> > Changes in nginx 1.0.5 19.07.2011
> > *) Feature: the referer_hash_max_size and
> > Referer_hash_bucket_size.
> > Thank you Witold Filipczyk.
> With these guidelines, there are some problems:
> - In the error messages as they are called, and referers_hash_max_size
> - The default value for referer_hash_max_size in the error message
> described as a 128 (actually - 2048)
> - It is not clear how to find exactly where the problem
> that is, such as the diagnosis is
> nginx: [emerg] could not build the referers_hash, you should increase
> referers_hash_bucket_size: 64
> But that's not enough, as the context for referer_hash_max_size server
> and location
> nginx mailing list
> nginx at nginx.org
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the nginx