A service needs attention after launch, too. Operational maintenance, also called production support, is the work of keeping it reliable for customers. The immediate concerns are avoiding downtime, maintaining acceptable application latency, and responding when something goes wrong. section 10.1.5
Plan support as part of delivery, rather than waiting for the first incident:
- Budget and optimize infrastructure resources so cloud or third-party costs do not grow unchecked. section 10.1.5
- Set up monitoring and alerts that reach the on-call team promptly. section 10.1.5
- Agree on service-level agreements (SLAs) that reflect how many customers a problem affects and how urgently it needs attention. section 10.1.5
- Keep a shared runbook of common customer problems and resolutions. Update it as new problems arise so later on-call teams can use what the current team learned. section 10.1.5
Table 10.2 Different severity levels for trouble tickets
|
Severity 1
|
This level occurs when the business is affected. Everyone should stop working on anything else and focus entirely on fixing this problem as soon as possible. The blast radius, in this case, can be that each company customer is affected. Examples: order drops or a security breach of customer data. SLA can be turnaround of four to six hours.
|
|
Severity 2
|
This level is critical and needs immediate attention, but the blast radius, such as the number of customers affected, may be smaller (with perhaps 40% of customers affected). Example: during the order checkout, the applying code option is broken, so any customer with a discount code cannot use it and hence needs to pay the total price-. This example isn’t a happy customer experience, right? SLA in this case can be 24 hours.
|
|
Severity 3
|
This level is not as critical as the preceding two. On-call support can take one to two days to triage the problem and get back. Example: an application needs to go through a security review before launch. Because the service has not been launched, no production effect occurs, so the SLA is two business days.
|
|
Severity 4
|
This level is not urgent and can wait. It has the least priority for the on-call team and will be addressed when trouble tickets of higher priority are addressed. Example: a marketing manager requests access to your internal platform to show sale numbers. In this case, SLA can be in days, such as five business days.
|
The severity table makes response priorities concrete: an issue affecting every customer calls for a different response from a checkout problem affecting a smaller group. Its turnaround times are examples for setting SLAs, not a substitute for agreeing on the needs of a particular service. section 10.1.5
Support also improves through practice. After a major incident, review its root causes without blame, choose corrective actions, and share skills across the team. Revisit on-call processes as requirements and technologies change. For a newly launched service, ask: who receives an alert, where will they find a resolution, and how will the team learn from the incident? section 12.2.6 section 10.2
