[bind10-dev] b10-auth stops responding for several tens of seconds after receiving IXFR

[Yoshitaka Aharen]

Hello, 

My colleague (familiar with DNS but not with BIND 10) tried to install
and configure BIND 10.
It retrieves jp zone successfully with IXFR.
About 1.3M domain names are registered under jp, but each zone update
contains very smaller number of updated records.

However, it takes about 4 minutes to finish zone update while BIND 9
takes only about 1 second.
Furthermore, b10-auth instances stop responding to queries for about 30
seconds while zone update. I think it is strange as b10-auth utilizes a
separate thread to (re)load zone. About 1 minute after b10-auth
instances resume to respond, they answer updated records.

BIND 10 is configured to use in-memory datasource with SQLite3 backend.
  "data_sources": {
    "classes": "IN": [
      {
        "type": "sqlite3",
        "params":
          {"database_file": "/usr/local/var/bind10/zone.sqlite3"},
        "cache-enable": true,
        "cache-zones": ["jp."]
      }
    ]
  }

The machine has 4 CPUs and there're 4 instances of b10-auth configured.
 $ ps ax | grep b10-au[t]h
 22385 pts/3    Sl   872:29 b10-auth
 22389 pts/3    Sl   890:43 b10-auth
 22390 pts/3    Sl   887:34 b10-auth
 22392 pts/3    Sl   889:52 b10-auth

Do you have any suggestion of configuration to improve zone update
performance?

Thanks,

-- 
Yoshitaka Aharen <aharen at jprs.co.jp>
Japan Registry Services Co., Ltd.