[AURORA-501] Improve HostMaintenance API

• Type: Story
• Status: Resolved
• Priority: Minor
• Resolution: Later
• Component/s: Maintenance, SRE
• Assignee: Unassigned
• Reporter: Joe Smith
• Created: 01/Jun/14 01:05
• Updated: 28/Dec/15 02:54
• Resolved: 28/Dec/15 02:54

Description

From tweingartner in response to https://reviews.apache.org/r/20285/:

The API seems a bit strange to me here. In particular, there is a HostMaintenance class, and methods:

HostMaintenance.start_maintenance(host_list)
HostMaintenance.end_maintenance(host_list)
...

Which all seems very repetitive to me. It might make more sense to have HostMaintenance's constructor take a list of hosts (already grouped) to operate on:

maint = HostMaintenance(possibly_grouped_host_list)
maint.status(host_subset=None) => return status of each host in possibly_grouped_host_list, or a subset thereof if given, raise hell if subset not in possibly_grouped_host_list
maint.start(host_subset=None) => return some sort of status of the operation, or nothing and just raise hell?
maint.drain(host_subset=None) => return next drained batch/host, do in a loop, like using a generator
maint.end(host_subset=None) => Usually given the output of maint.drain()

A usecase would then look something like:

maint = HostMaintenance([
  ['host1', 'host2', 'host3'],  # Batch 1
  ['host4', 'host5'],           # Batch 2 (followed by single hosts)
  'host6', 'host7', 'host8', 'host9'])
print "Current status:", maint.status()
maint.start()
for batch in maint.drain():
  for host in batch:
    # Do the funky chicken
  maint.end(batch)

Possibly having a HostMaintenanceController class to batch/coordinate/etc, using the HostMaintenance class to do this low level work, taking things like a grouping function, inter-batch delays, inter-host delays, "funky chicken" function to call for each host/batch, ability to monitor progress, possibly a "big red button, PANIC stop" ability.

Issue Links

• is related to: AURORA-43 Rename maintenance commands to have a consistent prefix