On 2025-02-26 01:58, Amos Jeffries wrote:
On 26/02/25 00:33, BENJAMIN DELANNOY wrote:
> Please detail what you mean by "choice" or "decision". For example, do
you want to stop the timer when Squid makes its final http_access
decision?
I mean I want to monitor the latency on what I could manage : the
communication between client & squid, the time squid uses to process
requests. But not the time spent by the destination server as I don't
manage it.
This could be multiple monitors if 1 is too complicated or not
possible. So yes for example a timer that stop when Squid makes its
final http_access decision (in case of http or bumped https) or its
final ssl_bump splice decision.
In theory what you want is the %busy_time logformat.
Please note that %busy_time
* includes the time spent on origin server-driven work (which Benjamin
may want to _exclude_) and
* excludes the time spent waiting for client I/O (which Benjamin may
want to _include_).
Alex.
However, the calculations for that are still apparently inaccurate. So
you would be best also logging these for comparison measures:
%dt = DNS latency
%<tt - %<pt = forwarding-latency
.. the time spent by Squid fetching the response.
.. excluding time waiting for it to arrive from the upstream server
.. also excluding anything that happens while the server is
delivering response (eg. for responses >4KB).
%tr - %<pt = total Squid latency
.. including delays from parallel traffic
.. excluding same as forwarding-latency above
HTH
Amos
_______________________________________________
squid-users mailing list
squid-users@xxxxxxxxxxxxxxxxxxxxx
https://lists.squid-cache.org/listinfo/squid-users
_______________________________________________
squid-users mailing list
squid-users@xxxxxxxxxxxxxxxxxxxxx
https://lists.squid-cache.org/listinfo/squid-users
