[bind10-dev] experience and problems setting up forwarding resolver

[Jeremy C. Reed]

I have been using the forwarding resolver since early December 
continually on my personal network. It has worked fine.

Today, I installed the latest release (a couple days old) on another 
machine.

This email documents my experiences and problems. If you recognize a 
Trac ticket for any of this, please let me know.

I installed to default prefix. b10-auth couldn't find its libraries. 
This Debian system had a /etc/ld.so.conf that had an include that 
referenced some other files. One listed /usr/local/lib but ldconfig -p 
didn't show it (even though the cache file was newer that the config 
files). So I ran ldconfig and then this worked.  I do not have this 
problem when installing to a custom prefix -- in that case the RPATH is 
defined, pointing to my libraries.

I used bindctl to disable the auth server and enable the resolver, 
but bindctl showed that the Auth module was still available. I didn't 
test configuring the Auth module then (b10-auth was not running).

I noticed an unclear error message:

> config set Boss/start_typo true   
Error: /Boss/start_typo
> config set Boss/start_resolver true
> 


It also showed the Resolver module. I was surprised that it was shown 
as I thought I had to restart cfgmgr. I forgot. (I think I posted wrong 
directions to bind10-users list regarding that.)

I configured the listen_on and tested with dig.

I next had this problem:

> config set Resolver/forward_addresses[0]/address "204.152.187.14"
> config set Resolver/forward_addresses[1]/address "204.152.187.17"
> config diff
{'Resolver': {'forward_addresses[1]': {'address': '204.152.187.17'}, 
'forward_addresses': [], 'forward_addresses[0]': {'address': 
'204.152.187.14'}}}

Notice the
	'forward_addresses': []

> config commit
Error: Error in config validation: Unknown item 
forward_addresses[0]Unknown item forward_addresses[1]
Configuration not committed

So then I tried:

> config set Resolver/forward_addresses [{ "address": "204.152.187.14" }]
> config diff
{'Resolver': {'forward_addresses[1]': {'address': '204.152.187.17'}, 
'forward_addresses': [{'address': '204.152.187.14'}], 
'forward_addresses[0]': {'address': '204.152.187.14'}}}
> config commit
Error: Error in config validation: Unknown item 
forward_addresses[0]Unknown item forward_addresses[1]
Configuration not committed

And so I did a "config revert" and tried again:

> config revert
> config set Resolver/forward_addresses [{ "address": "204.152.187.14" }]
> config diff
{'Resolver': {'forward_addresses': [{'address': '204.152.187.14'}]}}
> config commit
Error: Address must contain both the IPaddress and port
Configuration not committed

I was hoping it would use default port. (I know of problem not using 
spec, but I thought the code had hardcoded default too.)

So I added a port and that worked.

I tested it with dig and it worked. I exited bindctl. I copied 
existing resolv.conf and set new resolv.conf to point to 127.0.0.1.

I decided to run it as non-root. I created a user (bind10dns). I 
started bindctl and did:

> Boss shutdown
{
    "error": "Module 'Boss' not responding"
}


bind10 was now stopped, but the b10 processes were still running with 
init as the parent:

root     11641  0.0  0.1   5228  2980 ?        S    19:52   0:00 b10-resolver
root at bind10:/home/jreed# ps -p 11641 -o ppid
 PPID
    1

I manually killed msgq and everything stopped except cmdctl. So I 
killed it too.

Then I ran "bind10 --user bind10dns" and it said:

root at bind10:/home/jreed# [bind10] Error on startup: Unable to start 
b10-msgq: Unable to connect to c-channel after 5 seconds

I guess a permission problem but not identified.

I restarted bind10 as root (without --user bind10dns) so my DNS would 
work.

My config file was only readable by root. So maybe that was the 
problem. So I chgrp to my new group and chmod g+rw.

I ran bindctl and got:

> Boss shutdown
Error: 

What was the error? I don't know.

Now bind10 was not running and neither were the b10 processes.

Then:

> quit
Traceback (most recent call last):
  File "/usr/local/bin/bindctl", line 133, in <module>
    tool.run()
  File "/usr/local/lib/python3.1/site-packages/bindctl/bindcmd.py", line 
122, in run
    self.cmdloop()
  File "/opt/pkg/lib/python3.1/cmd.py", line 138, in cmdloop
    line = self.precmd(line)
  File "/usr/local/lib/python3.1/site-packages/bindctl/bindcmd.py", line 
271, in precmd
    self._update_all_modules_info()
  File "/usr/local/lib/python3.1/site-packages/bindctl/bindcmd.py", line 
265, in _update_all_modules_info
    self.config_data.update_specs_and_config()
  File "/usr/local/lib/python3.1/site-packages/isc/config/ccsession.py", 
line 365, in update_specs_and_config
    self.request_specifications();
  File "/usr/local/lib/python3.1/site-packages/isc/config/ccsession.py", 
line 360, in request_specifications
    specs = self._conn.send_GET('/module_spec')
  File "/usr/local/lib/python3.1/site-packages/bindctl/bindcmd.py", line 
232, in send_GET
    status, reply_msg = self._send_message(url, body)
  File "/usr/local/lib/python3.1/site-packages/bindctl/bindcmd.py", line 
224, in _send_message
    self.conn.request('GET', url, body, headers)
  File "/opt/pkg/lib/python3.1/http/client.py", line 918, in request
    self._send_request(method, url, body, headers)
  File "/opt/pkg/lib/python3.1/http/client.py", line 946, in 
_send_request
    self.putrequest(method, url, **skips)
  File "/opt/pkg/lib/python3.1/http/client.py", line 810, in putrequest
    raise CannotSendRequest(self.__state)
http.client.CannotSendRequest: Request-sent

I know why my bind10 was killed; it doesn't daemonize so when I closed 
terminal it was running in, it was killed -- but sometimes children 
didn't get killed. HUP or whatever signal was not trapped or passed to 
children?

Because my DNS was down, I couldn't login remotely since I was using 
Kerberos.  When back in I reverted by resolv.conf and just added my 
local 127.0.0.1 to the top.

I used nohup to start bind10 then.

Even with configuration readable, I still couldn't run as a different 
user. I also tried with making the /usr/local/var/bind10-devel/ owned by 
same group and readable and writable.

So currently running as root.