ResponseOps for Data Teams: The Harbour Street Weekend

Let me tell you a story about a data team that never existed, because it captures what ResponseOps feels like from the inside better than any definition can. Call them the Harbour Street analytics group: eleven people, a warehouse, a stack of ingestion jobs and a reputation for tidy pipelines. They were good at moving data. They had never rehearsed what to do when somebody else was moving it without permission.

The names, dates and numbers here are invented. The patterns, though, come straight from how the Ops Atlas treats incident response against attackers as a permanent capability, and I will point to its ResponseOps entry as we go. One thing worth saying upfront: the Atlas calls the word ResponseOps its own editorial label, not a recognised standard, so do not be surprised if your auditors have never heard it.

Friday, 16:40: the odd export

It starts with a small thing. An engineer notices that a scheduled export job has been writing a larger file than usual to a bucket outside the normal path. She mentions it in the team chat, somebody replies with a shrug emoji, and the afternoon moves on. Nobody says the word incident. Nobody writes down the time.

That missing timestamp turns out to matter enormously. The Atlas puts a lot of weight on the first practice in its list, which is declaring, because every later clock hangs off it. Several legal regimes count from the moment an organisation becomes aware of a breach, with the EU’s data protection board treating awareness as the point where you are reasonably sure an event has occurred. The Atlas lists a seventy-two-hour window under GDPR, a day for an early NIS2 warning, a four-hour limit for financial firms under DORA once an incident is classified, and a six-hour rule in India from CERT-In. The Harbour Street team had no idea which of these applied to them, and, worse, nobody owned the decision of when awareness had occurred. The Atlas names exactly that gap as an anti-pattern: a deadline missed because no one was assigned to make the call.

Saturday: well-meaning cleaning

On Saturday morning the on-call analyst did what many of us would. She rotated the warehouse credentials, deleted the suspicious export and shut the bucket. It felt decisive. It also removed three things the team would soon wish it still had: the original file, the access logs attached to the bucket’s old configuration, and any idea of what had been taken.

If you have ever fixed a flaky pipeline by deleting the failed run, you know the temptation. The difference is that a failed run is yours, while an intruder’s footprint is a crime scene. The Atlas recommends preserving first, collecting the short-lived material before the long-lived, and hashing whatever you gather so you can later show it has not changed. Even a modest team can do this with a shared folder and a two-line procedure.

For readers who live in pipelines, this overlaps with the discipline of tracking lineage. Their own article on how DataOps optimizes real-time analytics and stops pipeline breaks shows how much visibility a well-run data team already has, and that visibility is half of what a responder needs.

Sunday: the ransom note

On Sunday night a different message arrived: files in the lake were encrypted, and a note demanded payment. Panic is a poor project manager. Their first instinct was to restore from backups, and in the story they did, quickly and proudly, on Monday morning.

Two things went wrong. First, the backups they used had been taken after the intruder arrived, so the restore reintroduced the same encryption problem on a handful of hosts, a pattern the Atlas files under things to avoid. Second, the data that had left the building on Friday did not come back with the restore. The Atlas describes this as double extortion: you can recover your files and still face a person holding copies. Its myth list includes the belief that having backups turns the matter into a restore chore, and this story is the reason that item is there.

On payment, the Atlas keeps to what public authorities say. The sanctions office of the US Treasury advises against paying, while an early approach to police counts, in its words, as a significant mitigating factor. This is not advice from me or from the Atlas on what any particular company should do, and the legal side belongs with counsel. It is simply a reminder that the decision should be made deliberately, by named people, before a note arrives, not by whoever happens to be awake.

Monday: counting the clocks

By Monday lunchtime Harbour Street had finally gathered its leaders in one room. The first thing the room discovered was that the clocks had been running since Friday. The second was that nobody was sure whether the company counted as an essential entity, a financial one, or neither. Cross-border operations made it messier still. The Atlas offers a simple remedy that predates any crisis: sit down with counsel in advance and draw up a map of which obligations apply, who decides awareness, and who sends what to whom.

It also notes a few details that surprise people. A US securities rule counts four business days from the moment a company decides an event is material, not four calendar days. The healthcare rule it mentions allows up to sixty days. A US critical-infrastructure reporting rule has been written but, in the Atlas’s autumn 2026 snapshot, is not yet in force. The point of listing all of this is not to memorise it; the point is that a spreadsheet kept current beats a scramble.

What the team changed

After the dust settled, our fictional group did what good data people do and turned the postmortem into process. Here is the short list they ended up with:

  • A named responder lead and a named deputy, so the awareness decision always has an owner.
  • A one-page plan stored somewhere that does not depend on the warehouse, with phone numbers on paper.
  • A meeting channel outside company email and chat, because the Atlas warns against running response chatter on systems that might be compromised.
  • A write-once evidence folder, with a short card explaining how to hash and log what goes into it.
  • A restore rule: no backup is used until someone has confirmed it predates the first sign of trouble.
  • A twice-a-year tabletop exercise with a mixed audience of engineers, legal and communications staff.

The Atlas suggests a thirty, sixty and ninety day arc for exactly this kind of programme: leadership, a plan and out-of-band comms in the first month; playbooks for the likeliest scenarios and the clocks map in the second; an executive tabletop plus a handoff to whatever team runs incident command in the third. It stays honest about maturity, too, offering a five-step ladder from ad hoc to adaptive and cautioning that the ladder is its own proposal.

Measuring the improvement

Data teams love a metric, and the Atlas supplies a few that suit them. Dwell time is the gap between the earliest hostile activity and the moment you notice. For Harbour Street it would be counted from the odd export, though the true start was probably earlier. Then come time to declare, time to contain and how far the scope grew between your first and final count come next. The Atlas also reminds readers that tallying incidents tells you about the volume of work, not how well it was done, so a low tally should not reassure you.

If your estate spans several clouds, the question of where the evidence lives and who can reach it gets harder. A useful companion is this write-up on multi-cloud data operations for modern teams, handy when you ask which account holds what.

The friendly summary

If you take one thing from the Harbour Street story, let it be this: response is a skill you can practise, not a heroic act you improvise. Write down the time when somebody first suspects something. Preserve before you clean. Map your clocks before you need them. And use plain phrases in contracts, because the Atlas itself recommends the long-established words security incident response over its own label. The rest is rehearsal, and rehearsal is something data teams are already good at.

Related Posts

EndpointOps as a Data Problem: Why Three Device Lists Never Agree

Your console, your security sensor and your asset sheet each tell a different story about the fleet. Here is how to join them into one honest number.

Read More

Why Your Patch Coverage Number Is Wrong: A Data Team Diagnosis

Six symptoms data teams see when patching stalls or lies, each with a cause and a fix: silent agents, pending restarts, expired exceptions and window math.

Read More

IncidentOps for Data Pipeline Teams: A Seven-Step Runbook

A seven-step IncidentOps runbook for data teams: severity rules, roles, update cadence, safe paging and a 90-day rollout you can start this quarter.

Read More

Where SRE Stands Today and What It Means for Data Platforms

A grounded look at how site reliability engineering matured, where it is heading, and what data teams can borrow for their pipelines right now.

Read More

Eight Data Engineering Habits That Make ThreatOps Work

A data engineer’s first month inside a ThreatOps effort shows how freshness, lineage and honest metrics carry straight over from the warehouse.

Read More

VulnerabilityOps Data Audit: Ten Checks on a Findings Pipeline

A checklist-style audit that treats vulnerability findings like any other data product, with ten checks on feeds, joins, freshness, closure evidence and access.

Read More
Subscribe
Notify of
guest
0 Comments
Oldest
Newest Most Voted
0
Would love your thoughts, please comment.x
()
x