[squid-users] tcp_miss_aborted driving me crazy !

[Alex Rousskov]

On 08/21/2018 03:42 PM, --Ahmad-- wrote:

> the TCP_MISS_ABORTED seems down request and i dont know whats the
> wrong there

Assuming the second column in the sanitized access.log output below is
transaction response time, it looks like the client aborts (i.e. closes
the connection) if it receives no response from Squid for ~38 seconds. I
do not know why the server did not respond for so long or why Squid did
not forward the response to the client for so long. You may be able to
distinguish these two cases using a packet capture.


> 16:26:42  5575 ip1 36722 ip2 13071 TCP_MISS/200 425 GET ...
> 16:26:42  5673 ip1 52172 ip2 13022 TCP_MISS/200 425 GET ...
> 16:26:42  5673 ip1 43942 ip2 13026 TCP_MISS/200 425 GET ...
> 16:26:42  5673 ip1 49560 ip2 13010 TCP_MISS/200 425 GET ...
> 16:26:42  5673 ip1 37166 ip2 13001 TCP_MISS/200 425 GET ...
> 16:26:42  5673 ip1 47320 ip2 13015 TCP_MISS/200 425 GET ...
> 16:26:42  5707 ip1 60090 ip2 13016 TCP_MISS/200 425 GET ...
> 16:26:36 38905 ip1 38342 ip2 11807 TCP_MISS_ABORTED/0 0 GET ...
> 16:26:36 38884 ip1 39622 ip2 11718 TCP_MISS_ABORTED/0 0 GET ...
> 16:26:36 38843 ip1 40526 ip2 11852 TCP_MISS_ABORTED/0 0 GET ...
> 16:26:36 38781 ip1 42524 ip2 11893 TCP_MISS_ABORTED/0 0 GET ...
> 16:26:36 38724 ip1 48742 ip2 11552 TCP_MISS_ABORTED/0 0 GET ...
> 16:26:35 38656 ip1 55278 ip2 11913 TCP_MISS_ABORTED/0 0 GET ...

What logformat do you use? Did you reorder these records or were they
logged by Squid in the decreasing timestamp order?

The requests in each of the two groups above look concurrent to me. Do
you explicitly turn collapsed_forwarding on in your squid.conf? If you
do, does turning collapsed forwarding off solve this problem?


> as long as i can ping it from an iPV6 address

This is probably not an IP connectivity problem but a TCP- or HTTP-level
problem. You may want to try curl or wget to reproduce it without Squid.


HTH,

Alex.