ReleaseEngineering/How To/Create new slaves or move them to other pools

From MozillaWiki
Jump to: navigation, search

The intention of this page is to contain information about what to do when:

  • a new machine that has never existed before is added to the pool
    • for one of these reasons:
      • we've purchased more of the same host
      • we've re-purposed (e.g. Win7 slaves instead of WinXP)
  • the machine is being moved between pools (e.g. build pool to try pool)

Here's what IT needs to get done to help you (when a non-existent machine is being added):

  • DNS changes
  • re-labeling (DCOps)
  • attach dongles if the machines are running OSX

If you're replacing an Android panda go here

Here's what it takes to re-purpose machines or mobile boards on our side:

buildbot-configs puppet again changes? devices.json graph server slavealloc SSH keys
New host Yes TBD tegras/pandas both test&build machines inserts build machines
Re-purposed Yes TBD tegras/pandas N/A updates build machines


NOTE: Each host type might have Post-imaging steps: Setup freshly imaged slave

graph server instructions

  • login to cruncher and switch to the root account
  • ask for the gpg file that contains the credentials and double decrypt it
  • connect to the graphs DB to add the machines
  • build slaves post to the graph server some data after uploading
    • but you don't need to add individual machines IIUC since they post as a unique machine name ($OS_$branch)

slave alloc instructions

  • login to relengwebadn.private.scl3.mozilla.com
  • slavealloc.txt.gpg contains the credentials
  • /data/releng/www/slavealloc/slavealloc dbimport -D $db_url --slave-data mydata.csv
  • more details in adding a slave

staging machines

  • see bug 808043 if the slaves were staging machines and need the outbound traffic blocks to be removed