Update facil.io version (#3259)
* Link the framework test to the updated facil.io version
Reviewing the testing logs exposed the fact that facil.io detected 80
CPU cores on the host machine (?!)… which invoked 80 forks.
This large number of forks was never attempted before and it exposed a
number of issues which were fixed in the beta.2 release.
This release also includes other minor updates and fixes which are
probably irrelevant (such as updated to the pub/sub and a fix for the
Redis engine etc’).
* Bump facil.io version
* Update facil.io version
I doubt if this is necessary, but Michael mentioned there were issue
with the facil.io run (or, to be precise, that something happened on
the app server during, or just before, the facil.io run).
The updated version mainly focused on the Websocket client features and
minor refactoring, but the minor refactoring included changes to the
logic in the beta.2 fix for cluster mode, so maybe this will help.
I should probably point out that I stress tested the beta.4 today with
no adverse effects (no memory leaks, no crashes, all was good).
However, I have no way to test an 80 core cluster with remote clients
(I just don't have that kind of access), so I'm not sure if this fixes
whatever went wrong (if it's even facil.io related).
**Sidenote**: I removed the `0.0.0.0` binding from the command line.
The NULL address should support IPv6, while I fear the `0.0.0.0` might
limit the address scheme to IPv4 (I doubt it, but I never tested the
idea).
* Update to beta 5, for enhanced error handling for large clusters
* Limit the number of cores utilized to 7 worker cores ( + 1 sentinel)
With credit to Michael for point this out… at some point the added
processes don’t really help as much as they use up resources.