[nsd-users] IXFR regression in nsd 3.2.9?

[Stephane Bortzmeyer]

We upgraded to NSD 3.2.9 (from 3.2.8) because we encountered the
problem "Fix denial of existence response for empty non-terminal that
looks like a NSEC3-only domain (but has data below it)." (a nasty
problem with DNSSEC). But we now have IXFR issues.

On one name server, NSD 3.2.9 works fine, zones are IXFRed and work.

On another name server, with much more zones (and big ones), we
deleted the databases and compiled everything again with zonec (no
problem). The server works fine but, when a zone for which it is a
slave is modified, we see the following:

1) Zones are transferred from the master. nsd-patch -l says:

zone zonecheck.fr transfer id 4e4f serial 2010101202 timestamp
  1330954843.828433: seq_nr 0 of 287 bytes
zone zonecheck.fr transfer id 4e4f serial 2010101202: commit of 1 packets
   time Mon Mar  5 13:40:43 2012, from serial 2010101201, log message:
   xfrd: zone zonecheck.fr received update to serial 2010101202 at
   time 1330954843 from 192.134.4.2 in 1 parts

2) But the daemon does not take them into account (see the serial
number on ns2.nic.fr):

% check_soa zonecheck.fr
zonecheck.fr. 2010101202 (pri) dnsmaster.nic.fr.
zonecheck.fr. 2010101202 (sec) ns1.nic.fr.
zonecheck.fr. 2010101201 (sec) ns2.nic.fr.
zonecheck.fr. 2010101202 (sec) ns3.nic.fr.

We do not see in the log the message which indicates the update. With
3.2.8, we had:

Mar  5 13:20:38 ns2 nsd[25255]: Zone zonecheck.fr serial 2010101200 is
updated to 2010101201.

3) We also observe that nsd-patch now runs endlessly:

# nsdc patch
reading database
database region after loading domain names: 20487745 objects (20487745 small/0
large), 1327521688 bytes allocated (35924654 wasted) in 324102 chunks, 324101
cleanups, 9631752 in recyclebin 39547 0 0 0 0 4938 14951 23553 18949 19717 20350
20846 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0
database region after loading database: 63529025 objects (63529016 small/9 large),
2553031096 bytes allocated (37243126 wasted) in 623293 chunks, 623301 cleanups,
10819688 in recyclebin 0 29842 37070 0 21837 19073 27919 0 22414 25163 24075 0 704
815 720 0 32 152 160 0 45 49 52 0 36 42 51 0 1 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
reading updates to database
[Then nothing, until we kill the process]

  PID USER      PR  NI  VIRT  RES  SHR S %CPU %MEM    TIME+  COMMAND
11124 nsd       25   0 2496m 2.4g  256 R 100.0  7.7  11:27.74 nsd
11295 root      25   0 2472m 2.4g  668 R 100.0  7.6   8:11.07 nsd-patch

(With 3.2.8, it takes only 2-3 minutes on the same machine)

4) Downgrading to 3.2.8, everything works fine again.