rpcinfo Problem

[Volker Jahns]

Running rpcbind on a FreeBSD 6.1 testsystem has horrible effects, when

- rpcbind is started at system boottime by the rc.conf directive rpcbind_enable="YES"
- rpcinfo -p localhost is run ( this command then hangs until the system has died)


The top output shows high load and 'many' rpcbind processes which have been started.
--
last pid: 48637;  load averages:  3.99,  3.24,  3.23       up 0+07:47:18  16:02:42
1832 processes:3 running, 195 sleeping, 1633 waiting, 1 lock
CPU states:  5.2% user,  0.0% nice, 26.8% system,  4.3% interrupt, 63.7% idle
Mem: 121M Active, 20M Inact, 88M Wired, 4688K Cache, 34M Buf, 1004K Free
Swap: 470M Total, 244M Used, 226M Free, 51% Inuse, 22M In, 26M Out

  PID USERNAME  THR PRI NICE   SIZE    RES STATE    TIME   WCPU COMMAND
  317 root        1 128    0  1440K   424K RUN     44:17  7.37% rpcbind
37057 root        1  96    0  6524K  3468K RUN      0:20  0.06% top
  296 root        1  96    0  1300K     0K WAIT     1:06  0.00% <syslogd>
  437 root        1  96    0  3408K     0K WAIT     0:00  0.00% <sendmail>
  538 root        1  96    0  6092K     0K WAIT     0:00  0.00% <sshd>
  447 root        1   8    0  1312K     0K WAIT     0:00  0.00% <cron>
  441 smmsp       1  20    0  3300K     0K pause    0:00  0.00% <sendmail>
  383 root        1  96    0  1212K     0K WAIT     0:00  0.00% <usbd>
  541 root        1  20    0  3996K     0K pause    0:00  0.00% <csh>
99806 root        1   4    0  1468K     0K WAIT     0:00  0.00% <rpcbind>
38770 root        1   4    0  1464K     0K WAIT     0:00  0.00% <rpcbind>
20459 root        1   4    0  1468K     0K WAIT     0:00  0.00% <rpcbind>
21924 root        1   4    0  1440K     0K WAIT     0:00  0.00% <rpcbind>
  426 root        1  96    0  3356K     0K select   0:00  0.00% <sshd>
49102 root        1   4    0  1468K     0K WAIT     0:00  0.00% <rpcb>
10715 root        1   4    0  1468K   648K kqread   0:00  0.00% rpcbind
49102 root        1   4    0  1468K     0K WAIT     0:00  0.00% <rpcbind>
45921 root        1   4    0  1464K     0K WAIT     0:00  0.00% <rpcbind>
45947 root        1   4    0  1464K     0K WAIT     0:00  0.00% <rpcbind>
--

The output of some well-known commands w/ the system in this state is puzzling me:
--
orion# dmesg
No more processes.
--
--
ssh orion -l root
ssh_exchange_identification: Connection closed by remote host
--

Moreover, system log worries me:
--
Mar  8 08:20:26 orion kernel: kern.maxfiles limit exceeded by uid 0, please see
tuning(7).
Mar  8 08:20:26 orion kernel: kern.maxfiles limit exceeded by uid 0, please see
tuning(7).
Mar  8 08:20:26 orion syslogd: /dev/console: Too many open files in system: Too
many open files in system
Mar  8 07:20:25 orion rpcbind: warning: /etc/hosts.allow, line 23: cannot open /
etc/hosts.allow: Too many open files in system
--

Running rpcinfo -p <hostname> from a remote system can be used to benchmark this FreeBSD system. sockstat shows the TCP connects to rpcbind from the remote system and everything is fine.


If rpcbind is _not_ started at boottime, but from the commandline once the system is up, rpcinfo -p localhost works as expected.

I want to run NIS on the system, so rpcbind must run in reliable manner.

Any help is much appreciated.
-- 
Volker Jahns, volker at thalreit.de