Nils Goroll

FYI; I ran some trivial benchmark with `perf` profiling and the SMA locks are the main contender when we [remove the `private_oh` bottleneck](https://github.com/nigoroll/varnish-cache/tree/TRY_private_ohs_2). This is while running `wrk -c 1000...

> The thing I don't understand from prior discussions is why we can't use storage for synthetic responses just like we do with regular fetches. The current proposal is to...

This has been discussed during bugwash. Some questions/notes: There are two cases when `c_fail` of a `malloc` stevedore is increased: - [if the allocation would exceed the configured (net) amount...

Regarding the original issue of this ticket: You could have run into the issue if concurrent requests delivered objects with a total size exceeding the memory. The panic in the...

Re @waynebzhang Your report sounds similar to what has been discussed in this ticket and I have nothing to add to [my previous note](https://github.com/varnishcache/varnish-cache/issues/4408#issuecomment-3481560587). I will edit your post because...

See also 2a5345de14efee39838f87c57908fe10d02decc2: I would think we should either add the colon to all the "qualifyer" prefixes, or remove the check for colon

bugwash: match for either colon or space after the prefix

@pserrano For the simple case of your varnish cluster being stable (no nodes joining or leaving, health state always good), there is no relevant difference between the simple modulo hashing...

performance tests: ```vcl vcl 4.1; backend none none; sub vcl_synth { set resp.body = "42"; return (deliver); } sub vcl_recv { return (synth(200)); } ``` ``` /tmp/sbin/varnishd -a 127.0.0.1:8080 -f...

> I think we can have both. if we added, say, a `vcl_fail`, we should call that and that only for the "fast error" paths.