To subscribe to updates watch my GitHub repository. New posts get added as pull requests.

Scaling incident management for small and large organizations

31 Oct 2017

Introduction

This write-up is a way of managing incidents and outages (there are many other ways). An incident or outage can be anything which negatively affects productivity or costs a company’s revenue. Such as a website being down. I will mainly be discussing the process and not tools. I will assume the following tools are available.

  • An issue tracker (typically used to track work in an organization).

  • Communication channels (typically instant messaging chat or email mailing lists). Whatever the norms are for an organization to communicate internally to employees and externally to customers. Collaboration tools fall under this category.

Coordinating incident management may not always be straightforward for young organizations. In my experience, I’ve been a part of good incident management and also participated in bad incident management. This post is about some of my highlights for incident management. Here are some points to think about:

  • Teams within an organization should work together to help speed the resolution of an outage and help take stress off of the individual who is resolving the outage.

  • Assign roles to different people so that they can work asynchronous to each other where possible.

  • Adhere to the duties of the role. Concentration and focus can quickly and easily be taken away from resolving an incident. It helps to define roles, assign roles, and when a person is assigned a role they avoid going outside of their boundaries to help maintain focus to a speedy resolution.

Defining roles

Here are some ideas for roles which can operate independently by different people. Note: some people can take on multiple roles where it makes sense or based on the size of the team. The idea here is to scale these roles out to even major outages. Each role does not necessarily need to be comprised of people from the same team.

  • Incident Lead - Defines communication channels and engages stakeholders for appropriate action and updates. Also assigns roles when needed.

  • Technical Lead - The person leading the charge on fixing the problem. They’re in the terminal, looking at any logs, and actively resolving the outage.

  • Technical Assist - A support role and backup for the Technical Lead. Supports troubleshooting by assisting the Technical Lead when asked to do things. Perhaps, troubleshooting a different route than the Technical Lead without adversely affecting their efforts.

  • Scribe - Keeps minute notes for ongoing events and tags a timestamp with each event so that a timeline of events is documented.

  • Communication Lead - Fields all end user questions in chat channels and email. Summarizes regular status and updates to users every 30-60 minutes and communicates the news to users via a pre-determined communication channel.

  • Testers - Tests and validates solutions. Runs tests and gathers feedback when requested by the Technical Lead or Technical Assist. These should be separate people than the Technical roles because they have a fresh mind when approaching and validating the problem.

Duties by role

Incident Lead

The Incident Lead helps coordinate and organize the overall effort for managing the incident. Duties for the Incident Lead include:

  • Opens an incident issue if one doesn’t already exist. Classifies and identifies the priority of the incident depending on who is blocked and how. Add issue tags and links to related issues. Creates a shared chat channel where different team roles can get updates and collaborate.

  • Identifies stakeholders and engages them. Stakeholders being people involved with or affected by the issue. Once stakeholders are identified, then communication channels can be agreed on.

  • Defines communication channels. Open video conference meetings when necessary to communicate with higher ups or encourages pairing sessions where appropriate between other roles. Decides on the main point of contact channel for regular information updates by the Scribe to affected stakeholders. For example, a mailing list or a specific chat channel.

  • There should only be one main point of contact channel and all other places the initial incident is announced should be directed to that main point of contact channel for status updates.

  • If the team hasn’t organized into roles, then the Incident Lead should help with this transition. Ask the team who wants to be in what role or start assigning people to the role for them to perform the duties of that role.

  • When all appropriate roles have been assigned to people the Incident Lead documents this in the incident ticket (perhaps with a table).

Technical Lead

Duties for the Technical Lead include:

  • Reviews logs related to the outage.

  • Discusses with Technical Assist or other Engineers to help form a better picture of the outage.

  • Engages the Technical Assist to run off and troubleshoot or perform asynchronous technical tasks outside of what the Technical Lead is working on where it makes sense to parallelize on work.

  • Drafts a solution to the outage with the Technical Assist based on information gathered.

  • Implementor for the fix to the outage.

Technical Assist

Duties for the Technical Assist include:

  • Reviews logs related to the outage.

  • Discusses findings with the Technical Lead.

  • When requested, performs tasks assigned by the Technical Lead. If the Technical Assist cannot directly perform the assigned task, then it is the responsibility of the Technical Assist to coordinate with other people. This is important because it allows the Technical Lead to focus on their issue at hand.

  • For large teams, the Technical Assist is in charge of coordinating multiple people who would assist in tasks assigned by the Technical Lead or Technical Assist.

  • Ask the Technical Lead to bring them up to speed where necessary so they can continue to provide support and aide.

  • Communicates with the Scribe any updates when requested or even as events unfold. The Technical Lead should not be bothered with requests for status updates where possible.

Scribe

Duties for the Scribe include:

  • Keeps a timeline of notable events both internal and external to the team. Basically, lurk on the conversation between all roles in the team as well as pay attention to events outside of the team.

  • Notable events should be recorded with a timestamp of roughly when the notable event occurred. This does not have to be exact, but reasonably accurate.

  • At the end of the outage, the Scribe should draft a “Summary Timeline” of events only noting the most major and relative events of the outage.

  • If no notable events have been reported by the Technical Lead or Technical Assist within 30-60 minutes then the Script should actively ask the Technical Assist for a status update.

  • The Scribe communicates to the Communication Lead roles and updates them of notable events (or to state that there is no update since the last known update).

  • Based on the timeline of events, the Scribe should calculate some statistics for the incident. This helps the team measure their success when handling incidents. In the future, these metrics can help define a service level agreement for a service as well. Some recommended statistics:

    • If known, the total time between when an outage first occurred until when the outage was discovered. Also loosely known as the mean time to detect.

    • If known, total time when the outage first occurred until when the outage was resolved.

    • Total time when the outage was discovered until when the outage was resolved. Also loosely known as the mean time to resolution.

  • The Scribe must post in the incident ticket a comment including the following. A single comment is all that is needed.

    • Statistics for the outage.

    • A summary of the timeline of events (overview).

    • A full timeline of events.

Communication Lead

The Communication Lead can be more than one person depending on the stakeholders of an outage identified by the Incident Lead. This is mainly because there may be different audiences, which need to be updated regularly. Audiences may include:

  • Customers or users internal to the company.

  • Customers or users external to the company.

  • Upper management.

Duties for the Communication Lead include:

  • Do not communicate status updates more frequently than every 30 minutes. However, if there has been no communication to audiences within 60 minutes, then an update should be communicated.

  • When ready to communicate, the Communication Lead should reach out to the Scribe for the latest timeline of events, updates, and status relating to the ongoing incident since they last communicated to audiences.

  • Rather than communicating the timeline from the Scribe literally, the Communication Lead should massage the information from the Scribe and draft an appropriate message for the intended audience.

  • The Communication Lead should keep notes on when they communicate with audiences (a timestamp), what was said to the audience (a quote), and where the message was communicated (the channels used to field the communication). This will collectively be referred to as the timeline of communications.

  • When the incident is over, the Communication Lead should post a single comment on the incident issue the noting the timeline of communications. This is necessary because it serves as a reference in the incident. If the communication is too large to post as a single comment, then a separate “communication issue” should be created; the communication issue linked to the incident issue; and commented on the incident issue that the linked issue is for documenting the timeline of communications.

Managing an incident

A typical oversimplified timeline for an incident goes like this:

  1. An outage occurs or something which blocks people from working (or users from using).

  2. There is a delay between the outage occurring and somebody noticing there’s a problem. This may be an automated monitoring alarm going off or, in the worst case, a user reports the issue.

  3. Some troubleshooting occurs to discover a fix for the outage.

  4. The outage is resolved by implementing the fix.

While all of the above is going on, different audiences want to know what is going on depending on how they’re affected by the outage. Imagine trying to address the above timeline while also juggling the following:

  • People will ask for status updates. The longer an outage occurs, the more noisy audiences will be about wanting to know what is going on and staying updated.

  • Some people will want to know the root cause of the outage and ask about it. This root cause may not be directly related or relevant to implementing a fix for the outage.

  • Other people will want to help but don’t know how to contribute so they ask the person working on the outage.

All of the above points serve to distract and slow down resolving an incident. Which, prolongs the pain and stress felt by the individual who is working on a fix. It also prolongs the pain and stress felt by the audiences who are adversely affected by the outage.

When an incident is discovered the following should occur simultaneously:

  • An Incident Lead should be identified and engaged. The Incident Lead then takes over to help perform their assigned duties (such as assigning roles to other people). Usually the Technical Lead is engaged first when they’re the author, but they should immediately delegate to someone else being the Incident Lead so they can start focusing on a resolution.

  • The technical roles (Technical Lead, Technical Assist, and anyone coordinated by the Technical Assist) should work on the actual incident. Their focus should only be on resolving the incident as quickly as possible. It is the job of the other roles to remove outside distractions.

  • The Scribe lurks on the conversation of the other roles and performs their note taking duties. If necessary, the Scribe may engage the Technical Assist.

  • The Communication Lead handles internal and external communication to different audiences. They should only engage the Scribe for updates and status because they are recording noteworthy events.

At the end of the outage, the Testers are engaged to validate an outage is truly over and help bring a fresh perspective. This fresh perspective is important because they are less likely to make mistakes when validating since they weren’t directly involved with the implementation of fixing the outage.

After an incident concludes

After an incident is resolved, teams usually meet to discuss (not always the same day) the summarized timeline of events recorded by the Scribe. Evaluating events after an incident or outage is typically what is called a “postmortem review” in IT. The purpose of discussing this is to bring involved parties together to understand the following:

  • The root cause of the problem.

  • Attempt to come up with actionable feedback (also called action items) in which a team or organization can help prevent an outage for the same reason from occurring in the future.

  • Reviewing how the incident was handled via process. This is an opportunity to reflect and improve how incidents are managed going forward. Incident management can be refined over time as a team learns to work together on incidents.

Summary

By defining roles and assigning people to different duties the stress of managing an incident can be less of a burden on a single individual. It helps to balance the stress across multiple people so that a team can stay focused on resolving problems which block other people (like customers). Down time can affect organization reputation, customer confidence, engineering productivity, and can even be translated into cost for such losses (lost money).

It makes sense to try to formalize an organizational process (and even within a team) to resolve incidents as quickly as possible. Hopefully, I’ve helped your team or company get better at incident management if you decide to incorporate some of these ideas.


This article was last updated Nov 01, 2017.

Posted in Engineering with tags: Process, Tips


Comments: