• SummerOfCode2015
• FreeBSDTestClusterAutomation

FreeBSD Test Cluster Automation

• Student: Kamil Czekirda

• Mentor: Warner Losh

Project Description:

I would like to work on Testing Infrastructure (FreeBSD cluster).

My goal is complete infrastructure for automatization installs and tests of FreeBSD.

The base of this project will be iPXE and iPXE scripts as a backend to control nodes. For frontend I propose a small python webapplication.

The applications of this project are non-limited (testing kernels, network test, parallel packages building).

Current status

Howto

Scheme

TFTP

http://doc.freenas.org/9.3/freenas_services.html#tftp

Just put undionly.kpxe into main directory in your favourite server.

DHCP

Only add the condition:

if exists user-class and ( option user-class = "iPXE" ) {
    filename "http://192.168.22.3/menu.ipxe";
}
    else {
    filename "undionly.kpxe";
}

into your subnet section

and hosts, from my lab:

host node01 { hardware ethernet xx:xx:xx:xx:xx:xx; fixed-address 192.168.22.51; option host-name "node01"; }
host node02 { hardware ethernet xx:xx:xx:xx:xx:xx; fixed-address 192.168.22.52; option host-name "node02"; }
host node03 { hardware ethernet xx:xx:xx:xx:xx:xx; fixed-address 192.168.22.53; option host-name "node03"; }

NFS

http://doc.freenas.org/9.3/freenas_sharing.html#unix-nfs-shares

HTTP

I have two http services, one in standard 80 port, it's for static files, the iso and init menu.ipxe script. The second service is running on port 8080, it's small python webapp (management).

files in repository

  dhcp/   dhcp server example
  ipxe/   iPXE port
  master/ storage content
  mfsbsd/ mfsbsd for nodes
  tftp/   tftp server content
  www/    http server content

setup

• setup FreeNAS and jail
• install in jail: subversion, py27-sqlite3, rsync, curl, python, apache24
• start servers: dhcp, tftp, http and nfs
• create local copy of freebsd repo: svn checkout svn://svn.freebsd.org/base/head/ /cluster/usr/src/

• add node in management - http://192.168.22.3:8080/admin and set take_task.ipxe as boot

• add task (revision number)

workflow

• node starts with take_task status (take_task.ipxe as a boot)
• node sends DHCP request and gives an answer with next-server and filename options (filename "undionly.kpxe")
• node downloads undionly.kpxe by tftp protocol and execute it

• iPXE sends DHCP request and gives an answer with filename option: http://192.168.22.3/menu.ipxe

• menu.ipxe is iPXE script and asks the management for chainloading next script (chain http://${www}:${port}/menu/${net0/mac})

• management returns static file take_task.ipxe

• take_task.ipxe runs next chainloading: chain http://${www}:${port}/admin/take_task/${net0/mac}

• management creates new task by /cluster/new_task.sh script
• management changes node status from take_task to wait
• management chenges task status from new to preparing
• node waits for next chainloading
• when management finish preparing task change node status to cluster.ipxe
• node chainloads cluster.ipxe script and uses sanboot to boot mfsBSD
• mfsBSD mounts storage in /cluster
• mfsBSD runs /cluster/run.sh script
• run.sh script compiles the source, build world, kernel, make ftp and installs dist on hdd (current status is available in storage in nodeXX/run.log file)
• run.sh if everything goes fine changes node status to rebooting and boot to hdd.ipxe
• node reboots and changes boot status to take_task.ipxe

Bugs and not done

Schedule

The Code

svn repository

svnweb

Useful Links

Files

Weekly status report

• Weekly Status #1

• Weekly Status #2

• Weekly Status #3

• Weekly Status #4

• Weekly Status #5

CategoryHistorical