Hello knot,
I have recently started a long over due migration to knot 2.* and I have
noticed that the server.workers config stanza is now split into three separate
stanzas [server.tcp-workers, server.udp-workers & server.background-workers].
Although this is great for flexibility it does make automation a little bit
more difficult. With the 1.6 configuration I could easily say something like
the following
workers = $server_cpu_count - 2
This meant I would always have 2 cpu cores available for other processes e.g.
doc, tcpdump. With the new configuration I would need to do something like the
following
$avalible_workers = $server_cpu_count - 2
$udp_workers = $avalible_workers * 0.6
$tcp_workers = $avalible_workers * 0.3
$background_workers = $avalible_workers * 0.1
The above code is lacking error detection and rounding corrections which will
add further complexity and potentially lacking itelagence that is available in
knot to better balance resources. As you have already implemented logic in
knot to ensure cpus are correctly balanced I wonder if you could add back a
workers configurations to act as the upper bound used in the *-workers
configuration. Such that
*-workes defaults:
"Default: auto-estimated optimal value based on the number of online
CPUs or the value set by `workers` which ever is lower)
Thanks
John
_______________________________________________
knot-dns-users mailing list
knot-dns-users@lists.nic.cz
https://lists.nic.cz/cgi-bin/mailman/listinfo/knot-dns-users
