Incident Playbook: How to Plan for and Respond to Outages
January 10, 2017

Scott Klein
Atlassian

Share this

There's nothing like a major web outage to remind us how much our applications rely on other web services and technologies to function. In late October of last year, a Distributed Denial of Service (DDoS) attack on Dyn, one of the largest Domain Name Service (DNS) providers on the internet, disrupted service for consumer and business applications across the web. This attack shed light on the delicate interdependent nature of the web as productivity and uptime across the world was effected, even for organizations that didn't directly use Dyn as a DNS provider.

For an IT team, it can be a rude awakening when one major service provider's downtime triggers outages of technologies across the web, eventually hitting their application in its destructive path. As we saw with this attack, the disruption of one service can have a domino effect on many others. From an outsiders' perspective, it can make the internet seem like a wobbly house of cards. While you and your IT team might not see it that way, consider how an outage of this magnitude is perceived by your business colleagues and customers.

To alleviate these concerns, it's worth developing an incident response plan so that when another outage inevitably occurs, you are able to show confidence and transparency while managing the incident. Incident response plans can be developed in three steps to ensure you and your team are prepared to not only respond to an incident, but also maintain clear communications with stakeholders to quell panic.

1. Identify and quantify incidents

There doesn't have to be a major DDoS attack for a response plan to come in handy (think bugs in production, broken builds, login issues, etc.). Your team needs to be able to quickly identify and quantify the incident in order to appropriately respond.

If your team can answer these questions to define what constitutes an incident, they will be better equipped to respond:

■ How many customers are impacted?

■ How long does the incident need to last?

■ How degraded is the service and how do you determine severity level?

2. Create an incident communication plan

A well-built incident response plan is largely about effective communication, from who should be included in remediation conversations, to what is communicated to end users.

Consider the following when creating an incident response plan:

■ Identify the incident commander. This person is already at work or on call and available to provide direction during an incident.

■ Establish internal channels for alerting and communicating with teams reacting to the incident. We recommend creating a "war room," or dedicated chat room, to use for the duration of the incident and cut out unnecessary noise.

■ Know how you will communicate with end users. Whichever solution you use, be sure it's reliable and available when you need it. This usually means something that isn't hosted outside your core tech stack, because you don't want your communication forum having downtime when you need it most.

■ Build templates and preset language for common incidents. The last thing you need to worry about during an incident is how to phrase an alert during an already high-tension situation. While every incident is unique, most of them we find fall into a few broad categories: total outage, regional outage, delayed response times, to name a few. By deciding on some stock language for these templates and saving it ahead of time, you'll have more time to dedicate to the incident when it actually happens. This also accelerates the time from detection to communication.
Plan to send periodic updates during the incident to keep the team and stakeholders informed.

3. Communicate with transparency

During an incident, streamlined, transparent communications can free up your team to focus on fixing the problem while ensuring customers have the most up-to-date, accurate information. Even if your outage is the result of a third-party service provider, it's your responsibility to make sure your customers have the right information. While you don't want to play a blame game during an incident, there's nothing wrong with letting your users know that you're waiting on a fix from another provider. This helps your users understand that there isn't a lot you can do.

If the service provider having the outage is communicating with users somewhere, it's smart to direct your users there. During the Dyn outage, we saw a lot of Dyn's customers direct their users to Dyn's service page. This helped clear up the confusion around the interdependencies of these structures and helped users get the right information in the quickest possible way.

Postmortem and evolve

The lifecycle of an incident doesn't end when your application is up and running again, and neither should the actions that your team takes. Include these steps into your response plan to ensure you're covering all of your bases - both now and in the future.

■ After the dust has settled, draft an incident postmortem to explain what happened and what precautions are being implemented to prevent a repeat incident. It's always good to write a postmortem, especially for major outages. This is your opportunity to tell your customers what went wrong and what you're doing to avoid the incident in the future. A good postmortem includes an apology for any inconvenience your service outage may have caused and acceptance of responsibility. Finger pointing (even if it genuinely was another party at fault) is never a good look. Finish off your postmortem by explaining your remediation plan.

■ Conduct a post-incident review with your team to evaluate the processes and strategy enacted during the incident and how those can be improved to better handle future incidents.

There's no telling when the next incident will occur, what the scale will be or what systems will be affected, but if your team is prepared with a incident response plan, they can act swiftly to respond to the issue without the frenzy and panic that typically ensues. And when the next major domino drops and the rest of the world is furiously clicking "reload" and overloading customer service lines, your customers and stakeholders will receive clear, transparent communications the entire time.

Scott Klein is StatusPage Product Manager at Atlassian.

Share this

The Latest

February 24, 2017

Global revenue in the BI and analytics software market is forecast to reach $18.3 billion in 2017, an increase of 7.3 percent from 2016, according to the latest Gartner forecast. Gartner believes the rapidly evolving modern BI and analytics market is being influenced by the following 7 dynamics ...

February 23, 2017

An important aspect of performance monitoring is where the observer stands when looking at the IT scenario. Each participant has a different view of what is bad performance - network, database, web, system, user personnel, management and external people - customers, regulatory bodies etc. These are what I call viewpoints ...

February 22, 2017

An important aspect of performance monitoring is where the observer stands when looking at the IT scenario. If a complaint says the performance of an application is dreadful, the network man might say "Everything is fine" and the database man may agree, both saying "What's the problem?" All these people may say that the performance world is rosy but not to other people who have a different idea on what is rosy and what is not ...

February 21, 2017

Instapaper, a "read later" tool for saving web pages to read on other devices or offline, suffered an extensive outage 2 weeks ago. While Instapaper hit a unique problem — a file size limitation — its experience speaks to a much larger problem: scaling a database is difficult, and never quick. That basic fact explains why outages like this are surprisingly common ...

February 16, 2017

Hybrid Cloud is the preferred enterprise strategy, according to RightScale's 2017 State of the Cloud Report ...

February 15, 2017

IT departments often try to protect against downtime by focusing on the web application. Monitoring web application's performance helps identify malfunctions and their cause on a code level, so that the DevOps team can solve the problem. But, monitoring application performance only protects against application errors and ignores external factors such as network traffic, hardware, connectivity issues or bandwidth usage, all of which can have an impact performance and availability of a website ...

February 14, 2017

Everybody loves DevOps. In fact, DevOps is the hottest date in IT. That's because DevOps promises to satisfy the deepest longings of digital business — including fast execution on innovative ideas, competitively differentiated customer experiences, and significantly improved operational efficiencies ...

February 13, 2017

Forrester forecasted that direct online sales totaled 11.6 percent of total US retail sales in 2016, but digital touchpoints actually impacted an estimated 49 percent of total US retail sales, according to The State of Retailing Online 2017: Key Metrics, Business Objectives and Mobile report, released by the National Retail Federation’s Shop.org division and Forrester ...

February 10, 2017

Cisco's acquisition of AppDynamics – and the premium it paid – represents a "statement acquisition" that addresses several converging trends in both technology and financial markets. For strategic acquirers and tech investors, the acquisition is about delivering value to users and improving business outcomes through a go-to-market model that drives recurring revenues ...

February 08, 2017

Industrial and technological revolutions happen because new manufacturing systems or technologies make life easier, less expensive, more convenient, or more efficient. It's been that way in every epoch – but Continuity Software's new study indicates that in the cloud era, there's still work to be done ...