Nils Goroll
Nils Goroll
So what I think is happening here: vmod_dynamic creates _domain_ directors (basically resolver threads) also during `vcl_init{}`, the feature you are using which is vital to layering. These directors create...
@delthas does #4205 work for you?
Hi @delthas I spun another patch, which I hope now _really_ covers all the cases. But how can I be sure? Can you please give #4259 a try?
I am not against, but is this a generic solution to the problem? Would it not make much more sense to add some sort of cache partitioning where a certain...
I agree with both Walid's and Guiallaume's comments, but having this as a global flag really seems to be quite limited. Next up, someone might want `feature +disable_bans_for_labeled_vcls` or `+disable_bans_unless_on_admin_acl`....
@dridi per vcl feature flags sound like a very good idea. I'd like to keep the `CONNECT` discussion separate, but regarding BANs I think your proposal is sound!
bugwash: This needs additional implementation of per-vcl feature flags
bugwash consensus: wait for h2 patch which uses this to make a judgement if it can be avoided
So this looks very sensible to me. Just one question: Is it really the best option to keep `NULL` for "default probe", or should `NULL` become "no probe" and instead...
@dridi bugwash is go when https://github.com/varnishcache/varnish-cache/pull/4309#discussion_r2044014197 is addressed.