# Bug with fastcgi\_cache\_bypass

**URL:** <https://www.ruby-forum.com/t/bug-with-fastcgi-cache-bypass/193195>\
**Category:** NGINX\
**Created:** [August 12, 2010, 9:10pm UTC](https://www.ruby-forum.com/t/bug-with-fastcgi-cache-bypass/193195 "2010-08-12T21:10:14Z")\
**Posts on this page:** 2\
**Page:** 1

<div class="post-metadata">

**Author:** ![Matthieu\_T](https://www.ruby-forum.com/letter_avatar_proxy/v4/letter/m/71c47a/32.png) [@Matthieu\_T](https://www.ruby-forum.com/u/Matthieu_T)\
**Post date:** [August 12, 2010, 9:10pm UTC](https://www.ruby-forum.com/t/bug-with-fastcgi-cache-bypass/193195/1 "2010-08-12T21:10:14Z")

</div>

Hi,

I kept seeing “upstream sent unsupported FastCGI protocol version” in  
my error logs, and they went away when I turned fastcgi\_cache off.  
It turned out that the files holding the cache were corrupted,  
starting at random places instead of the FastCGI headers.

After trying to tweak things every possible way, I finally found out  
that the errors seemed to go away if I didn’t use  
fastcgi\_cache\_bypass.  
And later on, I realized that when using fastcgi\_no\_cache with the  
same variable as for fastcgi\_cache\_bypass, I don’t get any of those  
errors, and it works as intended.

I saw in the source that a warning was added recommending to use  
fastcgi\_cache\_bypass when fastcgi\_no\_cache is used, but not the other  
way around.  
Should they always be used together, or is this a bug ?

Thank you,

Matthieu.

–  
Matthieu T.

---

<div class="post-metadata">

**Author:** ![Matthieu\_T](https://www.ruby-forum.com/letter_avatar_proxy/v4/letter/m/71c47a/32.png) [@Matthieu\_T](https://www.ruby-forum.com/u/Matthieu_T)\
**Post date:** [August 12, 2010, 10:12pm UTC](https://www.ruby-forum.com/t/bug-with-fastcgi-cache-bypass/193195/2 "2010-08-12T22:12:20Z")

</div>

Hello!

On Thu, Aug 12, 2010 at 12:09:02PM -0700, Matthieu T. wrote:

> And later on, I realized that when using fastcgi\_no\_cache with the  
> same variable as for fastcgi\_cache\_bypass, I don’t get any of those  
> errors, and it works as intended.
> 
> I saw in the source that a warning was added recommending to use  
> fastcgi\_cache\_bypass when fastcgi\_no\_cache is used, but not the other  
> way around.

This warning is about fastcgi\_no\_cache semantics change  
(previously it was doing both \_no\_cache and \_cache\_bypass things).

> Should they always be used together, or is this a bug ?

Sounds like a bug.

Maxim D.