Duplicate alerts in UI

[bigon]

Hello,

Since I added multiple routes (with a continue: true) I see my alerts in double in the UI of alertmanager (version 0.20.0)

I find old bugs, but if I understand well that should be fixed already

Edit: If I remove the continue: true they are not duplicated

global:
  resolve_timeout: 5m
  smtp_from: '[email protected]'
  smtp_smarthost: 'smtpintern.example.com:25'
  smtp_require_tls: false

route:
  group_by: ['alertname']
  group_wait: 10s
  group_interval: 10s
  repeat_interval: 1h
  receiver: 'email.all'
  routes:
    - receiver: 'email.all'
      continue: true
    - receiver: 'web.Hangout'
receivers:
- name: 'web.Hangout'
  webhook_configs:
  - url: 'http://localhost:6000/create?room_name=Prometheusalerts'
- name: 'email.all'
  email_configs:
  - to: '[email protected]'
inhibit_rules:
  - source_match:
      severity: 'critical'
    target_match:
      severity: 'warning'
    equal: ['alertname', 'dev', 'instance']

[simonpasquier]

I agree it's confusing but at the same time, it's kind of expected since alerts matche the 2 groups. The UI could be improved to indicate that the groups are for different receivers.

Side-note, instead of using continue: true you can configure multiple integrations within one receiver like this:

route:
  group_by: ['alertname']
  group_wait: 10s
  group_interval: 10s
  repeat_interval: 1h
  receiver: 'receiverA'
receivers:
- name: 'receiverA'
  webhook_configs:
  - url: 'http://localhost:6000/create?room_name=Prometheusalerts'
  email_configs:
  - to: '[email protected]'

[bigon]

I could indeed use that, but now I expanded my route to the following and grouping the receivers like that would not work I guess

route:
  group_by: ['alertname']
  group_wait: 10s
  group_interval: 10s
  repeat_interval: 1h
  receiver: 'email.all'
  routes:
    - receiver: 'email.all'
      continue: true
    - receiver: 'web.HangoutPRD'
      match:
        environment: 'PRD'
    - receiver: 'web.Hangout'

[simonpasquier]

You can still avoid continue: true and avoid duplication with YAML anchors.

route:
  receiver: 'notPRD'
  group_by: ['alertname']
  group_wait: 10s
  group_interval: 10s
  repeat_interval: 1h
  routes:
    - receiver: 'PRD'
      match:
        environment: 'PRD'
receivers:
- name: 'PRD'
  webhook_configs:
  - url: 'http://localhost:6000/create?room_name=PRD'
  email_configs: &email-all
  - to: '[email protected]'
- name: 'notPRD'
  webhook_configs:
  - url: 'http://localhost:6000/create?room_name=Prometheusalerts'
  email_configs: *email-all

[bigon]

Thanks

Lille remark it seems that the . in email.all is making the parser unhappy

[simonpasquier]

@bigon right, example updated :)

[beorn7]

As my colleague @joe-elliott just ran into the same issue, I thought a bit about it:

It is confusing that the UI doesn't mention the receiver anywhere. You just see two completely identical groups because of that. Even the UI itself is confused by it: If I click the Info button on one group, both groups expand.

Perhaps it would be good to show the receiver somewhere. Or to dedup completely identical groups.

[simonpasquier]

Perhaps it would be good to show the receiver somewhere.

It was my idea too.

[bigon]

I discovered the other day that there is a drop-down menu (next to "Silenced" checkbox) to limit the receivers that are shown, this is definitely not obvious

[ipstatic]

We just hit this issue attempting to upgrade. The duplication did not happen with the version we are on now (0.12.0) with the "all" receiver selected. I agree that this is very confusing and think the grouping should be merged when the "all" receiver is selected.

[britcey]

I'll add to the choir - just spent way too much time figuring this out. Additionally, the 'Receiver' drop-down doesn't scroll down to show you all of the receivers - but you can type and it will auto-complete.

I recently implemented the routing technique @roidelapluie described in https://promcon.io/2019-munich/talks/improved-alerting-with-prometheus-and-alertmanager/ and this issue lead to a lot of head scratching.

I understand why now, but it's not obvious.

[leahoswald]

We just ran into the same issue. It is really really not obvious why there is a duplicate in this case. There should definitely be a hint to the receiver in the UI.

[hrdflt]

I can also confirm this is a problem for us. As others have said it is not obvious why this is the case. Looking forward to a fix!

[roidelapluie]

I am fixing this in #3289