Skip to content

Latest commit

 

History

History
95 lines (79 loc) · 4.15 KB

README.md

File metadata and controls

95 lines (79 loc) · 4.15 KB
Raw

cluster-preflight-check

cluster-preflight-check by its name targets to trigger cluster failures and verify some key configuration right before it goes to production. Cluster failure simulation by different people can go wild and the results can be different, even wrongly interpreted. This tool standardizes the steps as the simple CLI. It is carefully designed and must not change any configuration to harm the production.

Motivation

Convenient

Customers using tools like yast2-cluster/ha-cluster-bootstrap to setup a cluster. Before really pushing into production environment, it's necessary to make sure everything in cluster works well and is configured correctly. To archive above target, a lots of commands or steps will be included, like kill the primary process, check node/resource/cluster status, fence specific node, simulate split brain scenarios, show these commands' results and how to recover.
So it will be more convenient if we can integrate these commands into one tool.

Standard

Ship the standard commands.
Like how to kill process, how to check status, how to make split brain, there are lots ways to to that. In the cluster-preflight-check, we only provide one way to do this, the way we make sure it do works and is recoverable.

Reusable

Results can be reused.
cluster-preflight-check generates JSON result after each test case. The JSON result includes loggings, descriptions, timestamp, and whether this case has passed. It can easily be utilized by other tools/applications and be friendly with developer.

Traceable

Process can be traceable.
cluster-preflight-check has loggings and reports. For killing testcase, a report will be created. Report will includes test case description, action logging and explanation for possible results.

Recoverable

cluster-preflight-check will not bring new troubles or unrecoverable troubles after executing actions.

Features

Check environment

  • check hostname resolvable
  • check time service
  • check watchdog
  • check firewall

Check cluster state

  • check cluster service
  • check STONITH/Fence configuration
  • check nodes
  • check resources

Killing test cases

  • kill sbd daemon
  • kill corosync daemon
  • kill pacemakerd daemon

When running killing test case, a report will be created at /var/lib/ha-cluster-preflight-check;
Report will includes test case description, action logging and explanation for possible results.

Fence specific node

Make split brain

Others

  • JSON results for each test case
  • Python2.7 and Python3.4+ support
  • Tested in SLE12sp3, SLE12sp4 and SLE15sp1

Install

pip install cluster-preflight-check

Use

ha-cluster-preflight-check --help

usage: ha-cluster-preflight-check [-e] [-c]
                                  [--kill-sbd | --kill-corosync | --kill-pacemakerd | --fence-node NODE | --split-brain-iptables]
                                  [-l] [-y] [-h]

Cluster preflight check tool set

optional arguments:
  -e, --env-check         Check environment
  -c, --cluster-check     Check cluster state
  --kill-sbd              Kill sbd daemon
  --kill-corosync         Kill corosync daemon
  --kill-pacemakerd       Kill pacemakerd daemon
  --fence-node NODE       Fence specific node
  --split-brain-iptables  Make split brain by blocking corosync ports
  -l, --kill-loop         Kill process in loop

other options:
  -y, --yes               Answer "yes" if asked to run the test
  -h, --help              show this help message and exit

Log: /var/log/ha-cluster-preflight-check.log
Json results: /var/lib/ha-cluster-preflight-check/ha-cluster-preflight-check.json
For each --kill-* testcase, report directory: /var/lib/ha-cluster-preflight-check

Demo

  1. using cluster-preflight-check to check environment and cluster status
  2. using cluster-preflight-check to kill sbd process
  3. using cluster-preflight-check to kill corosync process
  4. using cluster-preflight-check to fence specific node
  5. using cluster-preflight-check to simulate split brain