Cortex: Difference between revisions

From ISRWiki
Jump to navigation Jump to search
(IP address)
 
(67 intermediate revisions by 2 users not shown)
Line 1: Line 1:
Cortex is a computation rack for VisLab humanoid robots. It contains 7 machines:
Cortex is a server used by VisLab for running simulations (not the ones with the iCub robot).
* 1 server that manages startup, shutdown and the file system of the clients;
* 6 clients (named <code>cortex1</code>...<code>cortex6</code>) that run user processes.
All clients numbered 1 to 5 mount the same file system. Therefore, performing changes in the file system of cortex[1-5] will reflect to all other four clients.
The client <code>cortex6</code> is separate for now, because it runs a 64 bit Linux.


''An older version of this article can be found at [[Cortex/Archive]].''
''Old information can be consulted at [[Cortex/Archive]].''


= Network setup =
= Specifications =


== Connectivity ==
As of 2017, there is one machine (cortex1) with these specs:
 
* 8 x [http://ark.intel.com/products/65523/Intel-Core-i7-3770K-Processor-(8M-Cache-up-to-3_90-GHz) i7-3770K] @ 3.50GHz processor
Cortex machines are connected to Cortex Switch, that links to VisLab switch with a fiber optic connection of 4Gbit/s.
* 16GB of memory (<code>sudo dmidecode --type 17</code> to see RAM speed and type)
 
* 112GB SSD drive + 1TB HDD drive
== Cortex nodes ==
* NVidia [http://www.geforce.com/hardware/desktop-gpus/geforce-gtx-670 GeForce GTX 670] graphics card (CUDA)
 
* internal ISR IP address: 10.10.1.1
Cortex server and clients have the following IPs and domain names:
* Server:  10.10.1.240, server.visnet
* Client 1: 10.10.1.1,  cortex1.visnet
* Client 2: 10.10.1.2,  cortex2.visnet
* Client 3: 10.10.1.3,  cortex3.visnet
* Client 4: 10.10.1.4,  cortex4.visnet
* Client 5: 10.10.1.5,  cortex5.visnet
* Client 6: 10.10.1.6,  cortex6.visnet
 
''For further details, see the [[Vislab#Network]] and the [[VisLab network]] articles.''
 
= Additional setup =
 
== Boot procedure ==
 
The clients boot via the network, using the PXE system. Each machine determines its own identity and asks the server for a kernel image and an initial ram disk. Kernel images and initial ram disks are stored on the server in the <code>/tftpboot/</code> directory. Kernels can be stock kernels, but the initial ram disk must be created in a way that enables booting from the network. This is not as bad as it sounds. It involves invoking the command <code>mkinitramfs</code>.
 
The server decides which kernel and initramdisk to send to each machine based on the information stored in the two files: <code>/tftpboot/pxelinux.cfg/default</code>(for cortexes 1-5) and <code>/tftpboot/pxelinux.cfg/01-00-19-d1-9e-e9-53</code> (for cortex6).
 
The roots and the homes of the users are also stored on the server machine, so they are mounted by each client at boot time.
 
=== Mounting of root directory ===
 
We are not sure which mechanism mounts the root filesystem, exactly. Here is the relevant line from <code>/etc/fstab</code>:
  # <file system> <mount point>  <type>  <options>      <dump>  <pass>
  /dev/nfs        /              nfs    defaults        1      1
 
=== Mounting of home directory ===
 
The home directory is mounted via the Upstart system a few seconds after booting. When rebooting the system, it is possible to login while <code>/home</code> is still not mounted. In that case, logout and login in order to load your environment variables correctly.
 
In November 2010, we created a file called <code>/etc/init/mountHome-net.conf</code> containing:
  description    "Mount network filesystems"
  start on started networking or runlevel 2
  exec /usr/local/bin/mountHome.sh
 
and <code>/usr/local/bin/mountHome.sh</code> containing:
  MOUNTED=$(mount | grep home)
  while [ -z "$MOUNTED" ]
  do
        su icub -c 'mount /home' &> /var/tmp/mountHomeUpstartOut.txt
        MOUNTED=$(mount | grep home)
        echo $MOUNTED
        sleep 1
  done
  ls /home > /var/tmp/mountHomeUpstartLs.txt
  runlevel > /var/tmp/mountHomeUpstartRunlevel.txt
 
After a successful mount, we should see something like
  $ mount | grep home
  10.10.1.240:/nfsroot.home on /home type nfs (rw,user=icub,noac,addr=10.10.1.240)
 
If <code>/home</code> is wrongly mounted with the <code>noexec</code> flag, users won't be able to execute binaries located inside it.
 
== Server machine ==
 
The server has:
* a boot folder for the clients at <code>/tftpboot/pxelinux.cfg</code>. It contains the files:
** <code>default</code> - default boot file;
** <mac_address> - specific for a machine with the given mac address.
* startup scripts for each machine at <code>/nfsroot/app</code>
 
== Client machines ==
 
The clients have:
* A superuser account (<code>compurack</code>) to administer system-wide settings (configurations, libraries, etc.)
* Normal user accounts. By default, the login script runs the contents of file <code>$HOME/.bash_env</code>, where users can set their environment variables, e.g., <code>export ICUB_ROOT=$HOME/iCub</code>. This works for both interactive shell sessions and non-interactive ones (i.e., commands remotely invoked by <code>yarprun</code>).
* A <code>yarp</code> account to update and install the YARP library. Variable <code>YARP_ROOT</code> is set by default to <code>/home/yarp/yarp2</code> for all users (in <code>/etc/bash.bashrc</code>) <-- change this policy
* An <code>icub</code> account with sudo privileges (created with <code>sudo adduser icub admin</code> on 2009-06-30) <-- change this policy
 
== System-wide libraries and repositories ==
 
=== YARP ===
 
yarp2 is installed for user icub, similarly to [[iCubBrain server configuration]].
 
For now, we don't use system-wide installation (<code>sudo make install</code>). We could use it again after we verify that a user can easily override global settings.
 
=== iCub ===
 
iCub is installed for user icub, similarly to [[iCubBrain server configuration]].
 
For now, we don't use system-wide installation (<code>sudo make install</code>). We could use it again after we verify that a user can easily override global settings.
 
=== Other libraries, manually installed ===
 
Please list here the system-wide libraries and applications that were installed by the superuser, especially the ones that do not have a clean 'make install' procedure but were manually installed into <code>/opt</code>:
 
* ARToolKit
* Ipopt-3.5.5-linux-x86_32-gcc4.2.4
 
CMake 2.6 does not come with the version of Ubuntu currently installed, but it is needed by the latest version of yarp, so we installed it via this [http://www.cmake.org/files/v2.6/cmake-2.6.4-Linux-i386.sh archive].
* cmake 2.6
 
=== Other libraries, installed with Ubuntu packages ===
 
These packages were installed with <code>apt-get install</code>
  libncurses5-dev
  libace-dev
  libgsl0-dev
  libgtk2.0-dev libgtkmm-2.4-dev libglademm-2.4-dev
  glew-utils libglew1.5-dev
  libglut-dev
 
OpenCV:
  libcv-dev libhighgui-dev libcvaux-dev
 
== User repositories ==
 
Each user should manage his own yarp2 and iCub repositories. <-- ''then we shouldn't have done sudo make install here :)''
 
We recommend to set your environment variables in a new file, called <code>~/.bash_env</code> and containing:
  export YARP_ROOT=~/yarp2
  export YARP_DIR=$YARP_ROOT/build
  export ICUB_ROOT=~/iCub
  export ICUB_DIR=$ICUB_ROOT/main/build
  export PATH=$PATH:$YARP_DIR/bin:$ICUB_DIR/bin
 
Refer to the [[RobotCub software]] article for further instructions.
 
= Other configuration =
 
== Subversion ==
 
We have set the following parameter in <code>/etc/subversion/config</code>:
  store-passwords = no
This implies that SVN will ask you for your password every time you do a commit. (Don't worry about changing your personal <code>~/.subversion/config</code> file: the parameter is not actually set there, so the global <code>/etc</code> setting is used.)
 
== Network tuning ==
 
  sysctl -w net.core.rmem_max=8388608
  sysctl -w net.core.wmem_max=8388608
  sysctl -w net.core.rmem_default=65536
  sysctl -w net.core.wmem_default=65536
  sysctl -w net.ipv4.tcp_rmem='4096 87380 8388608'
  sysctl -w net.ipv4.tcp_wmem='4096 65536 8388608'
  sysctl -w net.ipv4.tcp_mem='8388608 8388608 8388608'
  sysctl -w net.ipv4.route.flush=1
 
== Prompt ($PS1) ==
 
The prompt is set to <code>user@cortex?:pwd$</code> in <code>/etc/bash.bashrc</code>. With those settings, if you log in to Cortex1, the prompt will be <code>user@cortex1:~$</code>.
We chose to do so because sometimes it's convenient to have the number of the Cortex machine you're working on embedded in the prompt.
By default, though, this configuration is overridden in the users' <code>~/.bashrc</code> file, and the prompt is set to <code>user@source</code> regardless of the Cortex machine you log in to.<br>
If you want to inhibit this behaviour in <code>~/.bashrc</code> and thus have a prompt like <code>user@cortex?:pwd</code>, just comment these lines in your <code>~/.bashrc</code>:
  # set a fancy prompt (non-color, unless we know we "want" color)
  case "$TERM" in
  xterm-color)
      PS1='${debian_chroot:+($debian_chroot)}\[\033[01;32m\]\u@\h\[\033[00m\]:\[\033[01;34m\]\w\[\033[00m\]\$ '
      ;;
  *)
      PS1='${debian_chroot:+($debian_chroot)}\u@\h:\w\$ '
      ;;
  esac
However, for users created after 2009-05-07, the prompt is already set to <code>user@cortex?:pwd$</code> by default.
 
= Helper commands =
 
* Check the kernel: <code>uname -m</code>
* Check the file versions: <code>file</code>
* Set bash shell in <code>/etc/passwd</code>
* Check disk space: <code>du –sh /home</code>
* Check per-user processes: <code>ps -U <user></code>


[[Category:Vislab]]
[[Category:Vislab]]

Latest revision as of 13:59, 26 January 2018

Cortex is a server used by VisLab for running simulations (not the ones with the iCub robot).

Old information can be consulted at Cortex/Archive.

Specifications

As of 2017, there is one machine (cortex1) with these specs:

  • 8 x i7-3770K @ 3.50GHz processor
  • 16GB of memory (sudo dmidecode --type 17 to see RAM speed and type)
  • 112GB SSD drive + 1TB HDD drive
  • NVidia GeForce GTX 670 graphics card (CUDA)
  • internal ISR IP address: 10.10.1.1