An issue that was a warning that goes to critical doesn't send the resolution state #30

DaveWitchalls · 2016-08-10T15:52:38Z

I think this is semi by design, but any views appreciated.

We've come across this during testing. Using CPU use as an example;

If we set a warning threshold of 50%, an email gets sent. If it then goes to Critical at 70% the alert in PagerDuty is triggered. The problem for us is, if the CPU goes back to below the critical threshold, but still within the warning threshold, the resolution isn't sent PagerDuty as the incident is still live, even though not in a critical state.

Any thoughts or workarounds for this?

Thanks,
Dave.

majormoses · 2017-03-24T17:06:42Z

Hmm I am seeing the same and need to spend some time looking into this. I wonder if maybe we can work around using event suppression rules?

majormoses · 2017-03-24T19:16:03Z

Taking a look these are the only valid actions: https://sensuapp.org/docs/0.25/reference/events.html#how-are-sensu-events-created

create
resolve
flapping

I honestly have not played much with flapping in sensu but this might help with some of the situations but not all. By default a handler has handle_flapping to true. https://sensuapp.org/docs/0.24/reference/handlers.html#handler-configuration

majormoses · 2017-03-24T22:36:29Z

I checked on the pagerduty side and event suppression can only be done on initial ingest.

majormoses · 2017-03-24T22:41:30Z

@eheydrick what are your thoughts? This is not really pagerduty specific nor can I think of a pagerduty specific work around.

majormoses · 2017-03-27T13:50:50Z

I checked with pageryduty and event spurpession will not serve as a work around. Unless we want to resolve on state change (I am torn on this). I think we might want to pose a generic question to the sensu community and see what people think.

DaveWitchalls · 2017-03-27T13:57:45Z

Hi,

The workaround we had to go for was to include the resolved status in warning alerts. Far from ideal, but it does the job!

Dave.

majormoses · 2017-03-27T19:39:49Z

@DaveWitchalls did you apply this via a filter or mutator? I have not had any time to really look into this much and would like to try out your workaround and see if it works for us.

DaveWitchalls · 2017-03-28T15:46:22Z

Hi @majormoses

I asked the guy who did the work for me and got the below, it's reasonably long winded as it doesn't make much sense out of context, hope it helps!

NOTES ABOUT OUR USE OF SENSU

By default we only send alerts if the occurrences count reaches 5, however this is configurable using the "occurrences" check variable.
By default we send reminder emails if the occurrences count is divisible by 20, again this is configurable using a "remind_every" custom check variable.
PagerDuty handles it's own reminders/escalations.
We send all alerts (WARNING/CRITICAL/RESOLVED) via email.
We only send CRITICAL alerts for specific checks via PagerDuty (as defined by the pagerduty_alert_filter).

Therefore if an check state changes from CRITICAL to WARNING we use a mutator to generate a fake RESOLVE message to be sent to PagerDuty to clear the alert.
In this situation PagerDuty will send a RESOLVED alert which will show the new status text as WARNING not OK, an email will also be sent showing the status text as WARNING.

For checks which only require email alerts use the following handlers list:
"handlers": ["default", "mail_alert_handler", "mail_recovery_handler", "mail_resolve_on_warning_handler"]

For checks which require both email and PagerDuty alerts (critical only) use the following handlers list:
"handlers": ["default", "mail_alert_handler", "mail_recovery_handler", "mail_resolve_on_warning_handler", "pagerduty_alert_handler", "pagerduty_recovery_handler", "pagerduty_resolve_on_warning_handler"]

FILTER CONFIGURATION

/etc/sensu/conf.d/filters/alert_filters.json

NOTE the use of a custom key/value pair - remind_every - which is setup on the check. If not set it will default to emailing reminders every 20 occurrences.

{
  "filters": {
    "mail_alert_filter": {
      "negate": false,
      "attributes": {
        "action": "create",
        "occurrences": "eval: value == :::check.occurrences|5::: || value % :::check.remind_every|20::: == 0"
      }
    },
    "pagerduty_alert_filter": {
      "negate": false,
      "attributes": {
        "check": {
          "status": 2
        },
        "action": "create",
        "occurrences": "eval: value == :::check.occurrences|5:::"
      }
    }
  }
}

/etc/sensu/conf.d/filters/recovery_filters.json

{
  "filters": {
    "recovery_filter": {
      "negate": false,
      "attributes": {
        "action": "resolve",
        "occurrences": "eval: value >= :::check.occurrences|5:::"
      }
    },
    "resolve_on_warning_filter": {
      "negate": false,
      "attributes": {
        "check": {
          "status": 1
        },
        "action": "create",
        "occurrences": "eval: value == 1"
      }
    }
  }
}

HANDLER CONFIGURATION

/etc/sensu/conf.d/handlers/mail_handlers.json

{
  "handlers": {
    "mail_alert_handler": {
      "type": "pipe",
      "command": "handler-mailer.rb -s 'SENSU TEST'",
      "filter": "mail_alert_filter"
    },
    "mail_recovery_handler": {
      "type": "pipe",
      "command": "handler-mailer.rb -s 'SENSU TEST'",
      "filter": "recovery_filter"
    },
    "mail_resolve_on_warning_handler": {
      "type": "pipe",
      "command": "handler-mailer.rb -s 'SENSU TEST'",
      "filter": "resolve_on_warning_filter",
      "mutator": "mail_resolve_on_warning_mutator"
    },
    "mail_handler": {
      "type": "pipe",
      "command": "handler-mailer.rb -s 'SENSU TEST'"
    }
  },
  "mailer": {
    "admin_gui": "https://xxx.xxx.xxx.xxx/",
    "mail_from": "sensu@xxxxxxxxx.xxxxxxxxx.xxxxxxxxx",
    "mail_to": ["xxxxxxxxx@xxxxxxxxx.xxxxxxxxx"],
    "smtp_address": "127.0.0.1",
    "smtp_port": "25",
    "smtp_domain": "xxxxxxxxx.xxxxxxxxx"
  }
}

/etc/sensu/conf.d/handlers/pagerduty_handlers.json

{
  "handlers": {
    "pagerduty_alert_handler": {
      "type": "pipe",
      "command": "handler-pagerduty.rb",
      "filter": "pagerduty_alert_filter"
    },
    "pagerduty_recovery_handler": {
      "type": "pipe",
      "command": "handler-pagerduty.rb",
      "filter": "recovery_filter"
    },
    "pagerduty_resolve_on_warning_handler": {
      "type": "pipe",
      "command": "handler-pagerduty.rb",
      "filter": "resolve_on_warning_filter",
      "mutator": "pagerduty_resolve_on_warning_mutator"
    },
    "pagerduty_handler": {
      "type": "pipe",
      "command": "handler-pagerduty.rb"
    }
  },
  "pagerduty": {
    "api_key": "xxxxxxxxxxxxxxxxxxxxxxxxxxxxxx"
  }
}

MUTATOR CONFIGURATION

/etc/sensu/conf.d/mutators/mail_mutators.json

{
  "mutators": {
    "mail_resolve_on_warning_mutator": {
      "command": "/etc/sensu/mutators/mutator-mail-resolve-on-warning.rb"
    }
  }
}

/etc/sensu/conf.d/mutators/pagerduty_mutators.json

{
  "mutators": {
    "pagerduty_priority_override_mutator": {
      "command": "mutator-pagerduty-priority-override.rb"
    },
    "pagerduty_resolve_on_warning_mutator": {
      "command": "/etc/sensu/mutators/mutator-pagerduty-resolve-on-warning.rb"
    }
  }
}

MUTATORS

/etc/sensu/mutators/mutator-mail-resolve-on-warning.rb

#!/usr/bin/env ruby

require 'json'

module Sensu
  module Mutator
    class Mail
      class ResolveOnWarning
        def execute(input = STDIN)
          event = JSON.parse(input.read, symbolize_names: true)
          occurrences = event[:check][:occurrences]
          history = event[:check][:history]
          ### EDGE CASE CHECKS ###
          # Check if number of occurrences is > 20 (max history length) - if so change it to 20.
          occurrences = 20 if occurrences > 20
          # Exit if length of history < occurrences + 1 as it can't have been in an alert state beforehand.
          exit 1 if history.length < (occurrences+1)
          ########################
          test_array = Array.new(occurrences, "2")
          if history[0-(occurrences+1), occurrences] == test_array
            event[:action] = 'resolve'
            event[:check][:occurrences] = 1
            JSON.dump(event)
          else
            exit 1
          end
        end
      end
    end
  end
end

## Is called from Gem script. Program name is full path to this script
### __FILE__ is the initial script ran, which is
### /etc/sensu/mutators/mutator-mail-resolve-on-warning.rb
if $PROGRAM_NAME.include?(__FILE__.split('/').last)
  mutator = Sensu::Mutator::Mail::ResolveOnWarning.new
  puts mutator.execute
end

/etc/sensu/mutators/mutator-pagerduty-resolve-on-warning.rb

#!/usr/bin/env ruby

require 'json'

module Sensu
  module Mutator
    class PagerDuty
      class ResolveOnWarning
        def execute(input = STDIN)
          event = JSON.parse(input.read, symbolize_names: true)
          occurrences = event[:check][:occurrences]
          history = event[:check][:history]
          ### EDGE CASE CHECKS ###
          # Check if number of occurrences is > 20 (max history length) - if so change it to 20.
          occurrences = 20 if occurrences > 20
          # Exit if length of history < occurrences + 1 as it can't have been in an alert state beforehand.
          exit 1 if history.length < (occurrences+1)
          ########################
          test_array = Array.new(occurrences, "2")
          if history[0-(occurrences+1), occurrences] == test_array
            event[:action] = 'resolve'
            event[:check][:occurrences] = 1
            JSON.dump(event)
          else
            exit 1
          end
        end
      end
    end
  end
end

## Is called from Gem script. Program name is full path to this script
### __FILE__ is the initial script ran, which is
### /etc/sensu/mutators/mutator-pagerduty-resolve-on-warning.rb
if $PROGRAM_NAME.include?(__FILE__.split('/').last)
  mutator = Sensu::Mutator::PagerDuty::ResolveOnWarning.new
  puts mutator.execute
end

majormoses · 2017-03-28T18:08:05Z

@DaveWitchalls thanks for the info I will take a look and see if I can find some bastard amalgamation based on yours that works for us.

majormoses · 2017-03-28T21:01:46Z

interesting, though considering this I am not sure if its worth the effort for me right now to use a filter now that occurances is an extenstion: https://github.com/sensu-extensions/sensu-extensions-occurrences

portertech · 2017-07-11T15:07:04Z

Sensu events now have a "occurrences_watermark", the Sensu built-in "occurrence" filter now uses it instead of "occurrences" for the purpose of the resolve action. These changes are in the Sensu Core 0.29 release.

majormoses · 2017-07-11T16:17:56Z

ok, cool we can have a reasonable path forward when 0.29 is supported by this plugin...

kshep · 2017-09-26T15:43:48Z

ok, cool we can have a reasonable path forward when 0.29 is supported by this plugin...

@majormoses Any thoughts on whether/when that might happen? (for the record, Sensu skipped from 0.29 to 1.0.0 and subsequently 1.0.2 in July)

majormoses · 2017-09-26T16:47:34Z

@majormoses Any thoughts on whether/when that might happen? (for the record, Sensu skipped from 0.29 to 1.0.0 and subsequently 1.0.2 in July)

When someone is motivated enough and has the time to work on it. There are ~200 plugins and realistically 2 active maintainers (neither of us working for sensu) and we rely mostly on other community members to contribute.

Now beyond my boilerplate 🤷‍♂️ answer of when...I don't think it would be too hard taking a quick look at the plugin.

joshbenner · 2018-01-07T23:43:07Z

Does this plugin not work with Sensu >= 0.29? Or is there just work required to support changes related specifically to this issue?

majormoses · 2018-01-08T17:08:07Z

This plugin does work with all recent versions of sensu, I am currently running sensu 1.1.1 and do not have issues. The comment referenced is regarding a work around for moving from warning -> critical -> warning. The idea was to auto resolve the incident and create a new one based on occurrences_watermark. This is still a hack and I do see that pagerduty does potentially have a better solution now: https://www.pagerduty.com/blog/dynamic-notifications but I have not looked into that enough to know what the limitations are.

ashleyabrooks · 2018-01-11T20:52:47Z

Hello from the PagerDuty Team! Following up to confirm this is a limitation in the PagerDuty API rather than in the Sensu integration itself. At the moment, incidents are immutable so the parent incident can't be updated when the severity changes. You can click into the newest alert itself to get the latest data.

I've submitted a feature request from the maintainer of this integration so our product team knows mutable incidents are important to our customers. If you have any questions, please feel free to reach out to support@pagerduty.com.

majormoses · 2018-01-11T21:09:39Z

Thanks for confirming this.

joe-armstrong · 2021-01-25T16:05:05Z

I recently did some testing with this issue and it looks like PagerDuty has updated their API to allow for escalating an incident from a warning/low urgency to critical/high if the Dynamic notifications based on alert severity option is used in the Assign and Notify configuration for a service.

For testing we triggered an alert start as a warning starting a low priority incident in PagerDuty. I then changed the alert to the critical threshold which escalated the incident to high urgency in PagerDuty. One thing to note is that it will not de-escalate back to a low urgency incident.

majormoses added the Thinking label May 10, 2017

majormoses mentioned this issue Jul 11, 2017

Pagerduty alert text is not updated by handler #14

Open

mbbroberg added Status: Thinking and removed Thinking labels Sep 29, 2017

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

An issue that was a warning that goes to critical doesn't send the resolution state #30

An issue that was a warning that goes to critical doesn't send the resolution state #30

DaveWitchalls commented Aug 10, 2016

majormoses commented Mar 24, 2017

majormoses commented Mar 24, 2017

majormoses commented Mar 24, 2017

majormoses commented Mar 24, 2017

majormoses commented Mar 27, 2017

DaveWitchalls commented Mar 27, 2017

majormoses commented Mar 27, 2017 •

edited

Loading

DaveWitchalls commented Mar 28, 2017 •

edited by majormoses

Loading

majormoses commented Mar 28, 2017

majormoses commented Mar 28, 2017 •

edited

Loading

portertech commented Jul 11, 2017

majormoses commented Jul 11, 2017

kshep commented Sep 26, 2017

majormoses commented Sep 26, 2017

joshbenner commented Jan 7, 2018

majormoses commented Jan 8, 2018

ashleyabrooks commented Jan 11, 2018

majormoses commented Jan 11, 2018

joe-armstrong commented Jan 25, 2021

An issue that was a warning that goes to critical doesn't send the resolution state #30

An issue that was a warning that goes to critical doesn't send the resolution state #30

Comments

DaveWitchalls commented Aug 10, 2016

majormoses commented Mar 24, 2017

majormoses commented Mar 24, 2017

majormoses commented Mar 24, 2017

majormoses commented Mar 24, 2017

majormoses commented Mar 27, 2017

DaveWitchalls commented Mar 27, 2017

majormoses commented Mar 27, 2017 • edited Loading

DaveWitchalls commented Mar 28, 2017 • edited by majormoses Loading

NOTES ABOUT OUR USE OF SENSU

FILTER CONFIGURATION

/etc/sensu/conf.d/filters/alert_filters.json

NOTE the use of a custom key/value pair - remind_every - which is setup on the check. If not set it will default to emailing reminders every 20 occurrences.

/etc/sensu/conf.d/filters/recovery_filters.json

HANDLER CONFIGURATION

/etc/sensu/conf.d/handlers/mail_handlers.json

/etc/sensu/conf.d/handlers/pagerduty_handlers.json

MUTATOR CONFIGURATION

/etc/sensu/conf.d/mutators/mail_mutators.json

/etc/sensu/conf.d/mutators/pagerduty_mutators.json

MUTATORS

/etc/sensu/mutators/mutator-mail-resolve-on-warning.rb

/etc/sensu/mutators/mutator-pagerduty-resolve-on-warning.rb

majormoses commented Mar 28, 2017

majormoses commented Mar 28, 2017 • edited Loading

portertech commented Jul 11, 2017

majormoses commented Jul 11, 2017

kshep commented Sep 26, 2017

majormoses commented Sep 26, 2017

joshbenner commented Jan 7, 2018

majormoses commented Jan 8, 2018

ashleyabrooks commented Jan 11, 2018

majormoses commented Jan 11, 2018

joe-armstrong commented Jan 25, 2021

majormoses commented Mar 27, 2017 •

edited

Loading

DaveWitchalls commented Mar 28, 2017 •

edited by majormoses

Loading

majormoses commented Mar 28, 2017 •

edited

Loading