Underscore in handler name seems to be recognized as wildcard

nicolasbrechet commented 3 years ago

Expected Behavior

When defining several handlers with similar names, I want to name them like "myHandler_one", "myHandler_two", "myHandler_three". Then I want to define a check "myCheck_one" with handler "myHandler_one", another check "myCheck_two" with handler "myHandler_two", and so on.

Events from "myCheck_one" should be handled by handler "myHandler_one" only. Events from "myCheck_two" should be handled by handler "myHandler_two" only. ...

Current Behavior

All checks have all handlers!

"myCheck_one" has handlers "myHandler_one", "myHandler_two", "myHandler_three". Same for "myCheck_two" and "myCheck_three".

Possible Solution

It looks like the "" (underscore) character in the handler's name is seen as a sort of wildcard, where all characters after the are not taken into account.

Steps to Reproduce (for bugs)

Create multiple handlers with identical names, each with a different suffix containing an underscore, like "myHandler_one", "myHandler_two"
Create multiple checks, each with only one of these different handlers
Apply the configuration
Verify that all checks contain all the handlers

Context

We were trying to use several almost identical handlers depending on the priority, the suffix after the underscore was the priority. Example:

check_disk: email_priority1
check_cpu: email_priority2

Your Environment

Sensu version used (sensuctl, sensu-backend, and/or sensu-agent): sensuctl:6.4.0+ee, Sensu-backend: 6.4.0, sensu-agent: 6.3.0 and reproduced with 6.4.0
Installation method (packages, binaries, docker etc.): agents are installed with packages, sensu-backend running on kubernetes
Operating System and version (e.g. Ubuntu 14.04): agents are installed on windows 2012 R2, RHEL 7.9, CentOS 7.9,

amdprophet commented 3 years ago

Can you please include the handler & check configurations you used to produce this error?

nicolasbrechet commented 3 years ago

Of course!

Handler(s):

api_version: core/v2
type: Handler
metadata:
  name: email_P1
spec:
  type: pipe
    command: sensu-email-handler -f from@mail -s mailserver -i -P 25 -a 'none' -t 'to@mail' -S 'email subject' -T '/etc/sensu/template/email_template'
  timeout: 10
  filters:
  - not_silenced
  - is_incident
  - fatigue_check
  runtime_assets:
  - sensu-email-handler

other handlers were named email_P2, email_P3, etc and stored in configuration files email_P1.yaml, email_P2.yaml, ...

Check CPU:

---
type: CheckConfig
api_version: core/v2
metadata:
  name: check-cpu
  annotations:
    fatigue_check/occurrences: "5"
    fatigue_check/interval: "43200" # re-sends the alert every 12h.
    fatigue_check/allow_resolution: "true"
    email_priority: P2
spec:
  command: check-cpu.rb -w {{ .labels.cpu_warning | default 95}} -c {{ .labels.cpu_critical | default 99}}
  interval: 60
  publish: true
  handlers:
  - email_P2
  runtime_assets:
  - sensu-plugins-cpu-checks
  - sensu-ruby-runtime
  subscriptions:
  - system

After renaming all handlers without underscores, and updating the check configuration with the proper handler names, we no longer had the issue...

sensu / sensu-go