is there any guidance on using mod-rrl on a public server with a
moderate load, say 6kqps? we have rtfm, and remain unsure of
what we are doing. we want cookies, and therefore need to turn
rrl on. but with it turned on, we seem to drop a *lot* of
replies, a lot.
mod-rrl:
- id: default
rate-limit: 200
slip: 2
randy
I have static zones that are regenerated continually (every few
seconds) with a knotc zone-reload _zonename_ after.
Doing dig queries using ixfr= always used AXFR until I enabled per-zone
zonefile-load: difference
We saw logs periodically like:
warning: [foo.bar.example.com.] failed to update zone file (operation
not permitted)
This is because the files created are a different user than the knotd.
We are not doing automated signing nor dynamic updates.
This is with Knot 3.2.6.
1) Why would it try writing to the source zone file?
Anyways, we got rid of that warning with
zonefile-sync: -1
And the time to between the zone-reload and serving the new zone data
improved too.
I noticed with experimenting with journal-db-max-size and
journal-max-size the data.mdb was around 80% larger than the defined
size. As I removed and added records, the number of differences reduced
and I couldn't do IXFRs for many versions behind which I assume is
default journal-max-depth: 20. I believe this means the journaling was
dropping versions and does not require keeping at least 20 versions (not
a "minimum". I am fine with that.
I was able to trigger error like
error: [foo.bar.example.com.] zone event 'load' failed (not enough
space provided)
which I assume was due to journaling even when dropping old versions
still doesn't have enough space for most recent changes.
We are experimenting with increasing journal-max-size to work around
this.
It appears that the differences are fully handled in memory first before
updating to the journal file. I also see the journal file have large
jump in file size and never small increments.
2) Since I have no purpose to reuse a journal file for system restart or
recovery -- since I am building new zone files frequently, is there any
way to not use a journal file and just have knotd do this all in memory?
That way it is never writing to the journal every few seconds while I
continue to offer IXFR out support?
If not, maybe I will use memory file system for the journal files.
By the way, this is for around ten zones varying from around 1000
records to over 8 million records for the larger zones, They are
changing from a few records every few seconds to maybe a thousand
records every 30 seconds.
Thanks
