[squid-users] Website causing 3.5.5 squid crash

[Amos Jeffries]

On 7/06/2015 4:49 a.m., Mike wrote:
> Running Scientific Linux 6.6 (based on CentOS), compiled squid 3.5.5
> with no errors.
> Issue happens with both squid 3.5.5, and the newer 3.5.5-20150528-r13841.
> Starting squid service and running for a while, any attempt to access a
> website like www.nwfdailynews.com causes squid to crash and restart. The
> associated squid log entries from the same time squid crashes, seems to
> be several "TCP_SWAPFAIL_MISS" at that time:
> 
> 1433608644.053    117 192.168.2.110 TCP_SWAPFAIL_MISS/200 17246 GET
> http://launch.newsinc.com/77/css/NdnEmbed.css -
> HIER_DIRECT/184.51.234.134 text/css
> 

SWAPFAIL is either a disk I/O error, or the file has been erased from
disk without going through Squid making the cache index entry pointing
nowhere.

>From a service point of view thay are nearly harmless, though possibly a
sign that you need to check the HDD integrity.


> Other normal use including facebook and other website do not have any
> problems, so it has to be something relating to this site setup causing
> the crash. We are using this in a testing server for now but hoping to
> roll out to production level used by a few thousand people.
> 

That domain has bigger problems than what you see.

redbot.org shows its origin server is not generating Vary headers
properly. Which means visitors to the website that go through a proxy
(any HTTP proxy) will be getting randomly broken responses.

Furthermore, it appears to be served up to the public through a CDN
built from Squid proxies. So that is affecting *every* visitor.


Amos