On Apr 03, 2015, at 07:55, Martin Mosbeck <e0926555 [..] ...> wrote:

> Hi everyone,
>
> I have a rather interesting problem with the orocos deployment console:
>
> For my project I am deploying some components via the deployment console
> loading an .ops file. At the same time some ROS nodes are running. I
> tested the setup on 2 independent machines. They both have the same
> linux and orocos versions installed. One of the machines has a better
> CPU and more RAM and an SSD (compared to an HDD on the other machine).
> On the higher performance machine the following happens: whenever I do
> auto completion in the deployment console or search through the history,
> the deployer and the whole graphical system sometimes freezes for about
> 2 seconds and then resumes afterwards. On the other machine everything
> is running smoothly.
>
> My first thought was that maybe one of my components has a too high
> priority. My components are scheduled using ORO_SCHED_RT and the highest
> priority is 60. From my point of view 60 is not that high, and on the
> lower performance machine everything is running fine. I also think it
> should not matter that one machine has an SSD and the other an HDD. On
> both machines the realtime scheduling is enabled in /etc/pam.d/login
> and /etc/security/limits.conf .
>
> Any thoughts, why the deployer could freeze my graphical system on only
> one of the machines?
>
>
> Thanks,
> Martin

Any difference in the filesystem between the two machines? Are you running out of a NFS-mounted dir on one?

We have a very similar setup to you (include priorities and security settings), and see no issues (even with NFS-mounted dir’s).
S

Thank you for the quick answer.

Everything is local, no partitions are mounted via NFS. The only
difference is the filesystem type, but that should not be a problem, or
could it be? On the freezing machine everything is on a ext4 partition,
while on the other machine ROS and Orocos itself is on an ext4 and the
ROS nodes and Components are on an NTFS partition.

Martin

On 03.04.2015 14:13, S Roderick wrote:
> On Apr 03, 2015, at 07:55, Martin Mosbeck <e0926555 [..] ...> wrote:
>
>> Hi everyone,
>>
>> I have a rather interesting problem with the orocos deployment console:
>>
>> For my project I am deploying some components via the deployment console
>> loading an .ops file. At the same time some ROS nodes are running. I
>> tested the setup on 2 independent machines. They both have the same
>> linux and orocos versions installed. One of the machines has a better
>> CPU and more RAM and an SSD (compared to an HDD on the other machine).
>> On the higher performance machine the following happens: whenever I do
>> auto completion in the deployment console or search through the history,
>> the deployer and the whole graphical system sometimes freezes for about
>> 2 seconds and then resumes afterwards. On the other machine everything
>> is running smoothly.
>>
>> My first thought was that maybe one of my components has a too high
>> priority. My components are scheduled using ORO_SCHED_RT and the highest
>> priority is 60. From my point of view 60 is not that high, and on the
>> lower performance machine everything is running fine. I also think it
>> should not matter that one machine has an SSD and the other an HDD. On
>> both machines the realtime scheduling is enabled in /etc/pam.d/login
>> and /etc/security/limits.conf .
>>
>> Any thoughts, why the deployer could freeze my graphical system on only
>> one of the machines?
>>
>>
>> Thanks,
>> Martin
> Any difference in the filesystem between the two machines? Are you running out of a NFS-mounted dir on one?
>
> We have a very similar setup to you (include priorities and security settings), and see no issues (even with NFS-mounted dir’s).
> S