when access http://example.com/ its blank but working good with http://example.com/index.php i need to delete nginx cache file for http://example.com/ how to do it? Here my nginx cache conf.

    fastcgi_cache_path  /backup/cache levels=1:2 keys_zone=my-cache:1000m inactive=1000m;
    fastcgi_temp_path /backup/cache/tmp 1 2;
    fastcgi_cache_key "$scheme://$host$request_uri";

How to get the cache file name in /backup/cache folder?

  • Why do you also include $scheme:// in your cache-key? The scheme shouldn't make a difference with respect to the cached content, usually. May 30, 2019 at 14:01
  • @maxschlepzig HTTP could completely hide user information whereas HTTPS would show some info... Although the cookie should not travel along HTTP since that's a liability. Mar 7 at 17:03

6 Answers 6


one method, for example, but quite flexible an versatile

add this line to nginx.conf ...

proxy_cache_bypass $http_x_update;

... and you can anytime update any uri in cache with simple http request with "magic" header. for example

curl -s -o /dev/null  -H "X-Update: 1" mydomain.com


curl -s -o /dev/null  -H "X-Update: 1" mydomain.com/some/long/url/

for the security and satisfaction of paranoia :D you can change header name to any blablabla, for example

proxy_cache_bypass $http_x_gangnamstyle;
#proxy_cache_bypass $http_x_mycatsnickname;
#proxy_cache_bypass $http_x_b2ca678b4c936f905fb82f2733f5297f;


curl -s -o /dev/null  -H "X-GangnamStyle: 1" mydomain.com
curl -s -o /dev/null  -H "X-mycatsnickname: 1" mydomain.com
curl -s -o /dev/null  -H "X-b2ca678b4c936f905fb82f2733f5297f: 1" mydomain.com
  • 1
    Because i'm not using proxy, i changed to fastcgi_cache_bypass $http_X_Update; and then curl -s -o /dev/null -H "X-Update: 1" mydomain.com but still not working
    – imastar
    Mar 29, 2013 at 7:28
  • @imastar for fastcgi must fastcgi_cache_bypass $cookie_x_update directive. my answer slightly too late :) but can be useful to someone else
    – cadmi
    May 6, 2016 at 17:40
  • 5
    This doesn't delete anything, it just helps you to bypass the cache. Your users still would see the cached version. Also, that curl command already has "silent" flag why are you trying to save the output to "/dev/null"? that makes no sense.
    – xecute
    Sep 10, 2018 at 3:17
  • @xecute, the -s option just suppresses curl diagnostics - i.e. without -o /dev/null you still get the body printed to stdout (if any). Example: curl -s https://www.heise.de (compare this with the output when adding -o /dev/null to that call). May 30, 2019 at 13:43
  • This command does not delete stuff but it does INVALIDATE the cache. If you deleted the page from your site and want it to return a 404 - make sure nginx is set to cache 404 responses, otherwise you will keep geeitng the page
    – jitbit
    Jun 7, 2019 at 19:05

You must be having somewhere in your configuration this line,

fastcgi_cache_key *key*;

You need to find the key from there for your respective URL, and then calculate md5 for that key string.

Now suppose the md5 value comes to xm*****p3w, So, your cache file for the URL is /backup/cache/w/p3/xm*****p3w. Now delete it howsoever you wish.

The other automated way if you have nginx_cache_purge module with your nginx,

fastcgi_cache_purge CACHEREGION $cache_key;

where, CACHEREGION is the cache region defined by fastcgi_cache_path directive, and $cache_key is the value of fastcgi_cache_key directive.

Note: fastcgi_cache_purge directive is allowed in location block.

  • Nginx is not usually compiled with fastcgi_cache_purge. It comes with Nginx Plus
    – nicbou
    Jun 9, 2021 at 8:40

I created a simple script to find cache file location quicly and delete it:

#Usage sample : ~/cache_location.sh static.example.com/contents/3145.jpg
MD5=$(echo -n $1|md5sum|cut -d ' ' -f 1)
f1=$(printf $MD5 | tail -c 1)
f2=$(printf $MD5 | tail -c 3|head -c 2)
rm $CACHE_PATH/$f1/$f2/$MD5

It will calculate cache location based on Path and Filename. you need to install md5sum (yum install coreutils) .


Another potential solution, if you are looking for the actual file(s) in your file system would be to grep through the cache directory and print out the files one by one this way (assuming you're using the default /var/cache/nginx directory:

sudo grep -lr 'example.com/' /var/cache/nginx*

This should then print out the list of files to inspect and/or delete:


You can write a small shell function to compute the cache file location, e.g. for your levels=1:2 configured cache:

function nxcacheof {
    local x=$(echo -n "$1" | md5sum)
    echo "${x:31:1}/${x:29:2}/${x:0:32}"

Taking your cache-path and cache-key setting into account you can delete the cached entry for - say - https://example.org/foo like this:

$ rm /backup/cache/$(nxcacheof https://example.org/foo)
  • New to bash. Is the function something inside a .sh file? Can you show full example? Not sure how the $(...) is working in this context. Jan 12 at 9:11
  • @SteveTomlin You can copy'n'paste the function into your term to make it available. Or you copy it into a file xyz.sh and source it in your shell (source xyz.sh) to make it available. $() is command substitution syntax, i.e. the shell executes whats inside the parentheses in a sub-shell, captures its output and inserts it there. To see what's the result of the substitution you can also directly call nxcacheof https://example.org/foo in your shell. To get an idea for how this mechanism work you can also execute something like echo /backup/cache/$(nxcacheof https://example.org/foo). Jan 12 at 22:10

Create shell script to iterate over all found cache items for this domain



function purgecache {
  arr=($(grep -lr $1 /var/cache/nginx*))
  for val in ${arr[@]}; do
    rm $val
    echo "cleared cached for $val"

function main {
  purgecache "$@"

main "$@"

to call this method

sudo bash ./purgecache.sh mywebsite.com

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.