On 5/5/22 21:33, Manikandan Swaminathan wrote:
We're currently running Squid 4.8, and I want to know, what is the expected downtime when running "squid -k reconfigure"?
The actual delay depends on many variables, including the definition of "downtime". Very roughly speaking, reconfiguration today can be almost as heavy/slow (or as light/fast) as starting Squid from scratch. Certain startup actions (e.g., building an in-memory index of cache_dirs) are skipped during reconfigurations, and some optimizations (e.g., various internal caches) continue to work through reconfigurations, but a lot of heavy startup actions are performed during reconfigurations as well.
How does this affect existing and incoming connections?
Most existing connections, especially short-lived ones, are usually unaffected when the configuration does not change much. IIRC, new incoming connections may be rejected during reconfiguration (in some cases). The code is not written to guarantee much: Squid does _not_ maintain a consistent configuration state during (re)configuration. YMMV.
I ran a simple test in my machine where I reconfigure squid, while separately running multiple proxy requests. As far as I could tell, there wasn't any disruption, but I'd like to get some input from more experienced folks.
Your test did not expose existing problems. I do not know how much those problems are going to affect your deployment environment.
FWIW, we have an ongoing "smooth reconfiguration" project that makes Squid reconfigure just the changed configuration directives, without disrupting traffic. It will take a while to cover all configuration directives, but I hope to see the first pull requests that cover some directives soon.
HTH, Alex. _______________________________________________ squid-users mailing list squid-users@xxxxxxxxxxxxxxxxxxxxx http://lists.squid-cache.org/listinfo/squid-users
