Thank you very much for the quick response. It looks like that the cache now stores many small files because of the revalidation feature, and reached the inode storage limit.


Maxim Dounin Wrote:
-------------------------------------------------------
> Hello!
>
> On Thu, Sep 11, 2014 at 10:15:43AM -0400, arraisgabriel wrote:
>
> > Hi, recently we noticed that the version 1.7.3 added a feature
> important to
> > our infrastructure: "cache revalidation now uses If-None-Match
> header if
> > possible.".
> > So we changed part of our cache to the 1.7.4 version, but something
> strange
> > started to happen, at certain point of disk usage nginx started to
> return
> > 500 to all requests with this kind of message in the error log:
> >
> > [crit] 12908#0: *7209656 open() "/cache/nginx_tmp/0002938835" failed
> (28: No
> > space left on device) while reading upstream, client:
> xxx.xxx.xxx.xxx,
> > server: , request: "GET http://xxxxxxxx.net/ HTTP/1.1", upstream:
> > "http://xxx.xx.xxx.xxx:80/", host: "xxxxxxxx.net"
> >
> > It looked like that there wasn't enough space in the temporary cache
> > directory. But running df -h (and after sync && df -h) the result
> was:
> >
> > [user@nginx ~]$ df -h
> > Filesystem Size Used Avail Use% Mounted on
> > /dev/xvda1 7.8G 2.3G 5.4G 30% /
> > devtmpfs 3.7G 20K 3.7G 1% /dev
> > tmpfs 3.7G 0 3.7G 0% /dev/shm
> > /dev/xvdb 40G 12G 27G 31% 24% /cache
> > ---------------------------------------- used for cache
>
> ENOSPC from open() likely means you've run out of inodes, not disk
> space. Try looking into "df -i", it may be helpful.
>
> --
> Maxim Dounin
> http://nginx.org/
>
> _______________________________________________
> nginx mailing list
> nginx@nginx.org
> http://mailman.nginx.org/mailman/listinfo/nginx