wiki:GENIRacksHome/InstageniPlanetLab

Version 2 (modified by acb@cs.princeton.edu, 6 years ago) (diff)

--

InstaGENI-PlanetLab Status

This page describes the current state of PlanetLab nodes running on InstaGENI racks (a.k.a. InstaGENI-PlanetLab or IG-PL for short).

Deployment

We have allocated a few InstaGENI nodes through the Flack GUI and booted them as PlanetLab nodes. Following is a list of InstaGENI nodes currently running the IG-PL node image. Note that, as of this time, these nodes have not yet been opened up to the public; see the Still To Do section below.

  • pc3.instageni.cs.princeton.edu
  • pc3.lan.sdn.uky.edu
  • pc4.instageni.northwestern.edu
  • pc3.instageni.rnoc.gatech.edu

InstaGENI-PlanetLab Central

IG-PL uses a single Aggregate Manager to control all IG-PL nodes. This Aggregate Manager (which we call IG-PL Central) is running on node igplc.cs.princeton.edu. IG-PL Central is running the SFA and is federated with ProtoGENI and the GENI Clearinghouse. We have successfully used Flack and Omni to allocate slivers on IG-PL.

The MyPLC login page for IG-PL is http://igplc.cs.princeton.edu; currently only IG-PL admins can login.

  • URN: urn:publicid:IDN+igplc+authority+sa
  • URL: https://igplc.cs.princeton.edu:12346/
  • HRN: igplc.sa
  • igplc.gid:
    -----BEGIN CERTIFICATE-----
    MIICCDCCAXGgAwIBAgIBAzANBgkqhkiG9w0BAQQFADAQMQ4wDAYDVQQDEwVpZ3Bs
    YzAeFw0xMzA1MzAxNDUxNDhaFw0xODA1MjkxNDUxNDhaMBAxDjAMBgNVBAMTBWln
    cGxjMIGfMA0GCSqGSIb3DQEBAQUAA4GNADCBiQKBgQC0nm/cjhjwSses1NTBHuan
    MUvZEux0vpGfMEd/sX4nqhZwAanf108PM+UOwgK/oioYCXlYxeG1ycyALUGQ+3HS
    l6tPeYrFepNAY5IMIHLlTl4rWyUcqkvWrfgXM0k7grpfKC+p/MaUoZ0TPr5iB8B2
    gfjsx755pG09WPHnJRt1xwIDAQABo3IwcDAPBgNVHRMBAf8EBTADAQH/MF0GA1Ud
    EQRWMFSGI3VybjpwdWJsaWNpZDpJRE4raWdwbGMrYXV0aG9yaXR5K3Nhhi11cm46
    dXVpZDoxNmU5Njc0ZC00NzI1LTQ5MTMtODMwYS0zMDY4MzJjNTQ0NmEwDQYJKoZI
    hvcNAQEEBQADgYEARhbxiec6vLbeA0dGk+UeB/OCGCL3P+Q+wVj4HkWPqUyFMVVn
    0pqjTBo85+rfPYtS76g7CWcDm3YtnY4C23R9ct0482mbZTPSWcYVcMDASAKIU+45
    bZ8xLlFCWIAEqE8+oOXQiSBlclVS4A2+MzlXyeAOJzoNwx7Kkn3iD+AF8Qo=
    -----END CERTIFICATE-----
    

Currently on VICCI (http://www.vicci.org) we allow users to login with their PlanetLab passwords. When they do so, their existing PlanetLab slices are mirrored on VICCI so that they can easily add VICCI nodes to their PlanetLab experiments. If we decide that offering a similar sign-in for PlanetLab users on IG-PL would be attractive, we can easily enable this functionality.

Node Features

The nodes are running a recent PlanetLab build based on LXC and Open vSwitch, i.e., the same software as the VICCI testbed.

The software running on the nodes supports connecting containers to VLANs and GRE tunnel topologies as well as dedicated public IP addresses. However we need to do a bit more work to expose this functionality via the SFA and RSpecs. Currently containers created via the SFA can only connect to the public network via NAT (i.e., “PlanetLab” mode).

Right now the PlanetFlow traffic monitoring service is not running on the nodes. This will need to be fixed before these nodes are opened to users.

Integration with ProtoGENI

Below we describe some of the steps we took to integrate the PlanetLab and ProtoGENI models of managing nodes.

Installing PlanetLab

PlanetLab expects that a node is always booted from detachable boot media. The first challenge was to boot a ProtoGENI node from the PlanetLab Boot CD. One option we considered was to boot the node from a remote ISO file using the HP iLO’s Virtual Media function. Leigh and Rob convinced us that it would be better to boot from the hard drive, in order to not interfere with the normal operations of the InstaGENI racks.

Here is the procedure currently used to bring up an IG-PL node:

  • Boot an InstaGENI node using ProtoGENI’s standard Fedora 15 image
  • Once the node comes up, copy the PlanetLab BootCD ISO image for the node to /boot/iso/
  • Configure GRUB to use memdisk to boot the ISO directly
  • Use a modified PlanetLab installer to:
    • Create a new partition from the 986GB unused space on the disk
    • Install PlanetLab into the new partition without changing the existing ones

This method leaves us 848GB of disk space to host PlanetLab slices. A nice feature is that the node can then be “dual booted” as a ProtoGENI node (i.e., using the original F15 image) or a PlanetLab node simply by modifying /etc/grub.conf in the ProtoGENI partition.

ProtoGENI Node Status

In order to make the ProtoGENI node monitoring software see the PlanetLab node as "booted", it is necessary to do the following when the node boots (thanks to Leigh for this insight):

  • Mount the ProtoGENI partition on the PlanetLab node
  • chroot into the ProtoGENI partition
  • Run /usr/local/etc/emulab/tmcc.bin ISUP

Still To Do

  • Enable GENI-style networking in SFA. We need an example RSpec that has VLANs in it.
  • Bring up more nodes
  • Better integrate with ProtoGENI node status monitoring
  • Enable PlanetFlow on the nodes
  • Enable sign-on with PlanetLab credentials
  • Announce to PlanetLab users
  • Integrate with GMOC monitoring