I'm new to Unbound and in the process of deploying to our high-volume
production environment.
In order to meet some performance requirements, I have configured:
infra-cache-numhosts: 5000000
infra-host-ttl: 3600
Prior to restarting the service, I issued a stats_noreset followed by
dump_infra command.
The statistics output just prior to the restart showed 1.9m for
infra.cache.count
However, the dump_infra output only contained 470k lines. This was
true on 2 different servers with identical configurations.
Following the reload, I performed the same sequence of command at
around infra.cache.count 70k. The dump contained approximately the
same number of lines as the cache count.
The latter is what I had expected. However, this was not the case the
first time.
Can someone tell me what the expected results should be? Also, is
there a way to gets stats on size and count of the neg_cache ?
I'm new to Unbound and in the process of deploying to our high-volume
production environment.
In order to meet some performance requirements, I have configured:
infra-cache-numhosts: 5000000
infra-host-ttl: 3600
Prior to restarting the service, I issued a stats_noreset followed by
dump_infra command.
The statistics output just prior to the restart showed 1.9m for
infra.cache.count
However, the dump_infra output only contained 470k lines. This was
true on 2 different servers with identical configurations.
The dump_infra command does not print expired entries. Those are only
purged from memory when the limit (5m for you) is reached. (Expired
entries that are 'expired rto' are printed, those are backed off
unreachable servers).
So the missing lines are expired entries in the cache that have lazily
not been purged from memory yet. When you try that later on on a
freshly started server, the 1 hour ttl means that none have expired yet.