You can flush the Knot Resolver (kresd) cache on demand using socat to communicate with the running daemon:
socat - /run/knot-resolver/control/*
Inside the prompt type cache.clear()
> cache.clear()
You should get something like:
{
['count'] = 314182,
}
>
Type ctrl + C to exit.
If you run multiple kresd daemons. You might run into this error:
socat[990255] E exactly 2 addresses required (there are 18); use option "-h" for help
In this case, You can clear the cache on one instance, and it will clear for all the instances because they all share the same cache database (LMDB).