Skip to content

Why a small change creates a large workload

Why a small change can create a large workload: six small changes, each with why it is large and the source lines and sections that show it. The per-host work that every run pays, independently of restarts, closes the page.

Small change Why it is large Source
One option for nova-compute targets every compute; loop-notify adds libvirt + ssh restarts nova.yml:117-135; Tags §2.4; Reconfigure Nova §3
enable_<new service>: yes changes fluentd/logrotate on every host and nova.conf sections on every compute roles/common/defaults/main.yml:105-109, roles/common/tasks/config.yml:163-173; Reconfigure Nova §3
One extra controller changes the transport URL in every messaging service everywhere group_vars/all.yml:206; Change control-plane membership §2
A server-side ML2 option also merged into every OVS agent config roles/neutron/tasks/config.yml:172-190; Reconfigure Neutron §3
Rebuilt image of one service its siblings on the same host restart too library/kolla_docker.py:403-418; Tags §2.4; Update container images §2
Any OpenStack API service -t <svc> runs common, the service's haproxy include and Keystone registration even with no change; the DB sync runs on every deploy and, for most services, on every reconfigure (Neutron: deploy only by default). Infrastructure tags such as -t haproxy run no Keystone registration or DB sync Reconfigure Neutron §2, Reconfigure a single service §2; site.yml:126-392

And independently of restarts, every run pays the per-host work CODE. Per compute, a no-change -t nova run makes 46 remote module invocations (53 on compute001), and every full deploy or reconfigure makes 11 compare_container calls, counted from the task files in Compute scaling §2. The number of task results per compute is NOT AVAILABLE.

To remember

Lumo: if you keep only 5 things from this page, keep these.

  • One option for nova-compute is not small: it targets every compute, and loop-notify adds the libvirt and ssh restarts.
  • A rebuilt image of one service restarts its sibling containers on the same host too.
  • One extra controller changes the transport URL in every messaging service everywhere.
  • Every run pays the per-host work: every full deploy or reconfigure makes 11 compare_container calls per compute, independently of restarts. CODE
  • Write config first, apply in controlled batches: genconfig, then deploy-containers --limit batch by batch.

What would you do?

Lumo: a situation you could meet on a real platform, and three ways to react. Pick the one you would choose: I will tell you what this page says about it. Nothing is recorded.

A one-line nova-compute change

Your cloud has 40 computes. A ticket asks you to change one option for nova-compute. Your colleague says: "It's a one-line change, just run a full reconfigure now and close the ticket."

What would you do?

See every choice explained
  • Best choice: Write the config first with genconfig, then apply it with deploy-containers --limit, batch by batch.. One option for nova-compute targets every compute, and loop-notify adds the libvirt and ssh restarts. The page points to the pattern "Write config first, apply in controlled batches": genconfig, then deploy-containers --limit batch by batch.
  • Avoid: Run a full reconfigure of the cloud right away, as suggested.. The change is not small: one option for nova-compute targets every compute, and loop-notify adds the libvirt and ssh restarts. On top of that, every run pays the per-host work CODE: every full deploy or reconfigure makes 11 compare_container calls per compute.
  • Works, but risky: Run reconfigure on all computes in one go, but only at night.. The time of day does not change the size of the change: it still targets every compute, and loop-notify still adds the libvirt and ssh restarts on each of them. The page's pattern applies the config in controlled batches with deploy-containers --limit.

Check yourself

Lumo: 3 quick questions to check the main ideas of this page. Choose an answer to see the explanation and the section that gives the evidence. Your score counts your first choices and stays in this browser only.

1. You change one option for nova-compute. Why is it a large change?
Show the answer

Answer: It targets every compute, and loop-notify adds the libvirt and ssh restarts. First row of the table.

2. The image of one service is rebuilt. What else restarts?
Show the answer

Answer: Its sibling containers on the same host. Row "Rebuilt image of one service".

3. Does a run with no change cost anything per compute?
Show the answer

Answer: Yes: every full deploy or reconfigure makes 11 compare_container calls per compute. Every run pays the per-host work CODE, independently of restarts.