[squid-users] Cache index (swap.state) corruption when enabling/disabling SSLBump?

Amos Jeffries squid3 at treenet.co.nz
Sat Jun 20 00:21:26 UTC 2015


On 20/06/2015 9:46 a.m., Stanford Prescott wrote:
> I have a working SSLBump configuration with Squid 3.5.4. It seems that
> sometimes, if switching from HTTPS caching to only HTTP caching the cache
> becomes corrupted requiring deleting the swap.state file and recreating it
> by rebuilding the cache directories.
> 
> Has anyone else seen this? If so, has anyone found a fix that will prevent
> this from happening?

That is very strange. The cache has nothing to do with HTTPS vs HTTP.
All inputs and outputs are just URI+object with a hashe.

It dont matter if its http:// or https:// or ftp:/ or gopher:// or
whois:// ... its just a URI with a binary object attached.


I suspect you are probably encountering object corruption from in
incomplete shutdown/restart process. This can happen if there is a big
object in transit at the time, or if you set shutdown_timeout to a
small value, or used kill -9 to stop Squid.

Amos



More information about the squid-users mailing list