[Xorp-users] Error writing to routing socket: No such process

[Edwin Schokkenbroek]

On Jul 30, 2010, at 9:28 PM, Ben Greear wrote:

> On 07/29/2010 11:38 AM, Edwin Schokkenbroek wrote:
>> 
>> On Jul 28, 2010, at 11:35 PM, Ben Greear wrote:
>> 
>>> On 07/28/2010 01:12 PM, Edwin Schokkenbroek wrote:
>>> 
>>> 
>>> [ 2010/07/28 21:52:05.978695  INFO xorp_fea:2331 FEA fea/data_plane/io/io_ip_socket.cc:523 join_multicast_group ] Joined IPv4 group: 224.0.0.5 on interface epair3a vif epair3a  socket: 45
>>> [ 2010/07/28 21:52:05.985697  INFO xorp_rtrmgr:2330 RTRMGR rtrmgr/task.cc:2242 run_task ] No more tasks to run
>>> [root at router1 /]# [ 2010/07/28 21:52:07.70452  ERROR xorp_fea:2331 FEA fea/data_plane/fibconfig/fibconfig_entry_set_routing_socket.cc:527 delete_entry ] Error writing to routing socket: No such process
>>> [ 2010/07/28 21:52:07.70660  ERROR xorp_fea:2331 FEA fea/fibconfig_transaction.cc:80 operation_result ] FIB transaction commit failed on DeleteEntry4: net = 0.0.0.0/0 nexthop = 172.19.0.2 ifname = epair4a vifname = epair4a metric = 1 admin_distance = 1 xorp_route = true is_deleted = false is_unresolved = false is_connected_route = false
>>> 
>>> 
>>> Looks like it tries to remove a route that is already removed.  I just pushed a change
>>> that should cause that cause to not return an error.
>>> 
>>> Please grab the top-of-tree and see if that fixes the problem for you.
>>> 
>>> I don't have BSD system to build or test on, so let me know if you
>>> see any troubles.
>> 
>> The change seems to work. After compiling I now see the following messages, which seems good:
> 
> [snip]
> 
>> Is there a way to start the xorp_rtrmgr in a less verbose way ?
> 
> You can have it log all the messages to a file: -l xorp_log.txt

Well, that's very effective. thanks. xorp_rtrmgr starts very quiet now :-) Why didn't I think of this ?

> 
> You can also compile out tracing with several options, see 'scons --help'

I looked at the scons --help, as I'm not a programmer I didn't find an option which seems to do this. And from the output of scons I understand that most debug options are disabled:

Debug messages:    False
Debug function names:  False
Debug callbacks:   False

> 
> Might be a way to twiddle logging levels in the config file, but
> I'm too lazy to go look right now.
> 
> A lot of the more verbose stuff I put in there to provide more info
> if certain hard-to-reproduce bugs happen again, and the PFUNIX Listener
> stuff is already gone from top-of-tree, as it was breaking some
> regression tests.


Thank you for pointing this out, and providing and solving this in a very quick way. And did I already mention that Xorp is fun to work with ? 

Edwin





> 
> Thanks,
> Ben
> 
> 
> -- 
> Ben Greear <greearb at candelatech.com>
> Candela Technologies Inc  http://www.candelatech.com
> 


Edwin Schokkenbroek
edwin at spacebugs.nl

Engineers are full of great ideas. Making things better is what we're built to do. However, when it comes to engineers, there seems to be an inverse relationship between great ideas and great communication skills. This is what gets us into trouble.