With a fresh build of r30002 from trunk, on Ubuntu Hardy, I'm seeing periodic delays when connecting a GUI to a remote deployer. The suspicious bit is that the delays are almost exactly multiples of 5.0 seconds. I'm seeing this across multiple Orocos-based GUI/Deployer applications, some of which haven't been changed recently. Both the GUI and the deployer are run on the same machine, though they're pulling some data files across NFS (no errors shown there).
A sample orocos.log is attached. I haven't waited long enough to see if the GUI actually comes up.
Has anyone seen anything like this? Any ideas?
Cheers
Stephen
0.028 [ Info ][Toolkit] Loading Transport CorbaKDL. 0.028 [ Info ][Toolkit] Registered new 'CORBA' transport for frame 0.028 [ Info ][Toolkit] Registered new 'CORBA' transport for rotation 0.028 [ Info ][Toolkit] Registered new 'CORBA' transport for twist 0.028 [ Info ][Toolkit] Registered new 'CORBA' transport for vector 0.028 [ Info ][Toolkit] Registered new 'CORBA' transport for wrench 0.034 [ Debug ][Logger] Locating deployer named: LeftDeployer 0.034 [ Debug ][ExecutionEngine] Creating ExecutionEngine for NotFound 0.036 [ Debug ][ControlTaskProxy] ControlTaskProxy found CORBA NameService. 15.033 [ Info ][ControlTaskProxy] Successfully connected to ControlTaskServer 'LeftDeployer'. 15.033 [ Info ][ControlTaskProxy] Creating Proxy interface for LeftDeployer 15.033 [ Debug ][ControlTaskProxy] Fetching ScriptingAccess. 15.033 [ Debug ][ControlTaskProxy] Fetching Methods. 20.033 [ Debug ][ControlTaskProxy] Fetching Commands. 25.033 [ Debug ][ControlTaskProxy] Fetching Properties. 40.033 [ Debug ][ControlTaskProxy] Looking up Property stringFound valid assignment conversion from server string to local string 40.034 [ Info ][ControlTaskProxy] found!
Periodic delays in attaching to remote deployers
On Tuesday 03 March 2009 23:08:48 kiwi [dot] net [..] ... wrote:
> With a fresh build of r30002 from trunk, on Ubuntu Hardy, I'm seeing
> periodic delays when connecting a GUI to a remote deployer. The suspicious
> bit is that the delays are almost exactly multiples of 5.0 seconds. I'm
> seeing this across multiple Orocos-based GUI/Deployer applications, some of
> which haven't been changed recently. Both the GUI and the deployer are run
> on the same machine, though they're pulling some data files across NFS (no
> errors shown there).
>
> A sample orocos.log is attached. I haven't waited long enough to see if the
> GUI actually comes up.
>
> Has anyone seen anything like this? Any ideas?
It's TAO specific, and has been here since RTT 1.0. You won't have these
with OMNIORB. You can fix it by starting a Naming Service, and starting
your applications as described in the CORBA manual:
http://www.orocos.org/stable/documentation/rtt/v1.8.x/doc-xml/orocos-com...
The delay is caused by a five-seconds multicast(?) trial of the ORB to resolve the
initial reference of the naming service. Also 'wrongly' configured network interfaces
may cause trouble.
Why it only now shows up is not clear to me. I don't recall a patch that
enables naming for a code path that didn't do so before...
Are you using the Naming Service or passing IOR strings around ?
Peter
Periodic delays in attaching to remote deployers
On Mar 4, 2009, at 07:13 , Peter Soetens wrote:
> On Tuesday 03 March 2009 23:08:48 kiwi [dot] net [..] ... wrote:
>> With a fresh build of r30002 from trunk, on Ubuntu Hardy, I'm seeing
>> periodic delays when connecting a GUI to a remote deployer. The
>> suspicious
>> bit is that the delays are almost exactly multiples of 5.0 seconds.
>> I'm
>> seeing this across multiple Orocos-based GUI/Deployer applications,
>> some of
>> which haven't been changed recently. Both the GUI and the deployer
>> are run
>> on the same machine, though they're pulling some data files across
>> NFS (no
>> errors shown there).
>>
>> A sample orocos.log is attached. I haven't waited long enough to
>> see if the
>> GUI actually comes up.
>>
>> Has anyone seen anything like this? Any ideas?
>
> It's TAO specific, and has been here since RTT 1.0. You won't have
> these
> with OMNIORB. You can fix it by starting a Naming Service, and
> starting
> your applications as described in the CORBA manual:
>
> http://www.orocos.org/stable/documentation/rtt/v1.8.x/doc-xml/orocos-com...
>
> The delay is caused by a five-seconds multicast(?) trial of the ORB
> to resolve the
> initial reference of the naming service. Also 'wrongly' configured
> network interfaces
> may cause trouble.
>
> Why it only now shows up is not clear to me. I don't recall a patch
> that
> enables naming for a code path that didn't do so before...
>
> Are you using the Naming Service or passing IOR strings around ?
I use the naming service as per Orocos' instructions. Always have.
Turns out that Gnome's lovely Network Manager had mangled my hosts
file. }:-(
Omniorb is starting to look better and better ... if only I can find
time to play with it ...
Sorry for the noise.
S
Periodic delays in attaching to remote deployers
With a fresh build of r30002 from trunk, on Ubuntu Hardy, I'm seeing periodic delays when connecting a GUI to a remote deployer. The suspicious bit is that the delays are almost exactly multiples of 5.0 seconds. I'm seeing this across multiple Orocos-based GUI/Deployer applications, some of which haven't been changed recently. Both the GUI and the deployer are run on the same machine, though they're pulling some data files across NFS (no errors shown there).
A sample orocos.log is attached. I haven't waited long enough to see if the GUI actually comes up.
Has anyone seen anything like this? Any ideas?
Cheers
Stephen