If I chose to do nothing, the answer was I would be left of orphaned hosts in Redhat Satellite or Spacewalk and over time this could lead to higher license costs if leveraging support for Redhat Linux or just pure database bloat due to having all these previously used instances still referenced in my database.
This issue bothered me and I wanted a mechanism that would cleanup instances one they were terminated but the question was how to go about it?
Well I soon realized that OpenStack keeps a record of all the instances it ever created and or terminated. It was the terminated part that would be a key component to what I wanted to accomplish. I figured if I could mine out the deleted data of instances, I could cross check those against Redhat Satellite or Spacewalk.
The Perl script below does just that. I have it run every 24 hrs out of cron and it first goes into the OpenStack nova database and scrapes the instances table for any instances that were marked deleted in the last 24 hours. Any instances it finds it puts into an array that I then enumerate through using the spacecmd tools and check within Satellite or Spacewalk to see if the host is registered. If the host is registered, I then remove the host given that it is no longer a valid host that is up and running.
#!/usr/bin/perl
$cmd=`rm -r -f /root/.spacecmd/spacewalk.schmaustech.com`;
$yestdate=`TZ=CDT+24 /bin/date +%y-%m-%d`;
#$yestdate=`TZ=CDT /bin/date +%Y-%m-%d`;
chomp($yestdate);
@delhosts=`mysql -e "select hostname,uuid,deleted_at from nova.instances where deleted_at is not null"|grep $yestdate`;
foreach $delhost (@delhosts) {
($hostname,$uuid) = split(/\s+/,$delhost);
$uuid2 = $uuid;
$uuid2 =~ s/-//g;
@cmdout=`spacecmd -q system_details $hostname.schmaustech.com`;
foreach $cmd (@cmdout) {
chomp($cmd);
if ($cmd =~ /$uuid2/) {
$message = "Removing from satellite hostname: $hostname with UUID: $uuid...\n";
$cmdtmp = `logger $message`;
$cmdtmp = `spacecmd -y -q system_delete $hostname.schmaustech.com`;
}
}
}
exit;
