Johan Bergström johan at
Wed Mar 25 20:36:38 MSK 2009

On Mar 25, 2009, at 18:16 , Igor Sysoev wrote:

> On Tue, Mar 24, 2009 at 07:22:03PM +0100, J?r?me Loyet wrote:
>>>> Other question, how can I determine the maximum number of cache  
>>>> entrie
>>>> I have with a particular conf ?
>>>> proxy_cache_path  /data/nginx/cache  keys_zone=one:10m;
>>>> What is the size of keys on shared mem (X bytes). So that I can
>>>> extrapole the maximum number N of cache entries with YMo of shared
>>>> memory (N = (Y x 1024 x 1024) / X). (if there is nothing elese in  
>>>> the
>>>> shared mem and if the key size is constant)
>>> The key size is 128 bytes.
>> Thanks for the answer.
>> Is there a way to know the effective used size in the SHM segement ?
> No, if shared memory is overflowed you will see in error_log the  
> message:
> "ngx_slab_alloc(): failed". The problem is that currently no way to  
> see
> what memory zone is overflowed: no name and subsystem (proxy cache,
> ssl_session_cache, limit requests, etc.) I'm going to fix this.

Would it be possible to somehow add this into the status module, so  
it's possible to graph both disk allocation and other stats such as  

>> Another question about proxy_cache_valid. Let say I set it to 10m. I
>> do some hits and the requests are cached well. Now I update my conf
>> and set proxy_cache_valid to 30s and send a HUP signal to nginx. How
>> is already cached files timeout handled at restart or reload ?
>> Is the cache going to be mark as stale 30s after ?
>> Is the cache going to stay in the cache until its 10m timeout is
>> reached (and then be recached with a 30s timeout) ?
> The second. The valid time is stored in both memory and file as
> now + proxy_cache_valid. Therefore, a response will be valid
> for 10m after its creation, and then will be recached for 30s.
> -- 
> Igor Sysoev

More information about the nginx mailing list