Closed sbraz closed 2 days ago
I just saw 51176640603b123a38b49ae271cb9c9de9df1d28 which seemed to fix a very similar problem. Could it be a new regression?
Not really a regression. This is accurate from Varnish's perspective of successfully "sending" bytes to the kernel, but what happens beyond that is out of our control.
There's probably non-portable room for improvement, but this is the current state of affairs.
As @dridi explained. Otherwise: PRs welcome
Expected Behavior
Hello,
The doc for %O says:
So if bytes were not sent, I do not expect them to be counted.
Current Behavior
Currently, %O and %b always return the size of the entire response, even when the connection is interrupted by the client after a fraction of seconds.
Possible Solution
No response
Steps to Reproduce (for bugs)
varnishncsa -F "%b %O"
curl -m 0.5 http://localhost/some/very/large/file
Operation timed out after 500 milliseconds with xxx out of yyy bytes received
Context
I am currently attempting to count bytes sent to clients to determine which IPs are responsible for most of the traffic and I would like to have a reliable of doing this by analyzing Varnish's logs.
Varnish Cache version
varnishd (varnish-7.1.1 revision 7cee1c581bead20e88d101ab3d72afb29f14d87a)
Operating system
Debian 12
Source of binary packages used (if any)
https://packages.debian.org/bookworm/varnish