Atlassian stopped selling Opsgenie on 4 June 2025. Support ends on 5 April 2027, after which the product is switched off and data that has not been migrated is deleted. Existing customers may renew until then and were pointed towards Jira Service Management and Compass. As I write this on 5 September 2026, that leaves seven months.
This is not a complaint about Atlassian. Retiring a product it acquired is a legitimate commercial decision, it was announced with roughly two years of notice, and there is a documented path off it. It is a useful case study precisely because the vendor behaved reasonably. The migration still happened to people who had not planned it, on a date they did not choose, for a system whose job is to tell them production is broken.
The migration you did not schedule
An on-call platform looks small from the outside. It holds schedules, rotations, overrides, escalation policies and a handful of teams. Any of that can be rebuilt in an afternoon.
The schedules are not the problem. The integration keys are. Every alert source in your estate points at the vendor: Prometheus Alertmanager receivers, CloudWatch alarm actions, Azure Monitor action groups, Datadog monitors, Grafana contact points, a Lambda somebody wrote in 2021, a cron job on a box nobody has logged into since, and a webhook URL pasted into a third-party vendor console by a colleague who left two years ago.
You can find most of them with a grep across your infrastructure repositories. You cannot find the ones that are not in a repository, and those are the ones that will page nobody at 03:00 on the first weekend after cutover. The real work of a forced migration is inventory, not rebuild.
Four costs, only one of which appears in a budget
- Engineering time. Inventory, rebuild, dual-run, cut over, then chase the stragglers for a month. This is the cost everybody estimates, and it is the smallest of the four.
- Licence overlap. You pay both vendors during the parallel run, because turning the old one off before you trust the new one is not an option in this category.
- Risk during cutover. Every other migration can be rehearsed against a staging environment. This one cannot fully be, because the failure mode you care about is "an alert that should have paged somebody did not", and that is invisible until it matters.
- Negotiating position. You have none. The deadline is public, every vendor in the category knows it, and the discount you might have won in a normal renewal cycle is not on the table when the alternative is 5 April 2027.
The bill for the obvious move
The obvious move is PagerDuty. PagerDuty lists Business at $41 per user per month on annual billing, $49 billed monthly, with Enterprise quoted by sales (pagerduty.com/pricing, checked 5 September 2026). At 300 responders, Business is $147,600 a year.
Opsgenie Standard, for comparison, lists at $19.95 per user per month billed annually — a renewal rate for existing customers rather than something you can buy today (atlassian.com/software/opsgenie/pricing, checked 5 September 2026). At the same 300 responders that is $71,820.
So the forced migration roughly doubles the annual cost of being reachable by telephone, from about $72,000 to about $148,000, and you are making that decision against a deadline you did not set. That is what vendor risk costs. Not an outage — a number.
What ownership would and would not have changed
Owning the system would not have made the work disappear. Someone still runs it, patches it, and carries the pager for the pager. Anyone who tells you ownership is free is selling something.
What ownership changes is who sets the date. Nobody sends you an end-of-support notice for your own repository. If you decide to rewrite the notification service, you do it in a quarter with capacity, not in the quarter Atlassian chose.
For reference, Klaxon — our on-call implementation — runs at about $640 a month on AWS at 300 responders, roughly 40,000 alerts and 450 incidents a month, with the notifier in two regions. That is $7,680 a year, plus carrier charges for SMS and voice passed through at cost, which run around $180 a month at that alert volume. Call it $9,840 a year all in. The build is nine weeks with two engineers, including import of PagerDuty or Opsgenie schedules, escalation policies, services and integration keys.
The part of that estimate that matters most is the last clause of the build note: existing webhook URLs keep working behind a compatibility shim. A migration where you do not have to touch four hundred alert sources is a fundamentally different project from one where you do. If you ever do this — with us or without us — build the shim first. It converts an inventory problem into a routing problem, and routing problems have deterministic answers.
Three questions to ask of every SaaS system you run
Do this once a year, for an hour, per material vendor. It is the cheapest risk work available to a technology function.
- If this vendor gave us 24 months notice tomorrow, is that a nuisance or an emergency? Write down which. The emergencies are your real dependency list, and it is usually shorter and stranger than the architecture diagram suggests.
- How many credentials pointing at this vendor exist outside our own repositories? If the answer is "we would have to ask around", that is the number that makes a migration expensive.
- What does the export actually contain? Not what the documentation says is exportable — run it, and look at the file. In the Opsgenie case, data not migrated by 5 April 2027 is deleted, so the export is the only artefact that survives.
Where the honest answer is still to keep buying
Paging is a category with real irreducible dependencies. You do not escape the carriers: SMS and voice still go through a telephony provider, and running your own is not a serious suggestion. You do not escape the app stores if you want native push on a phone that is in do-not-disturb. And if the reason you bought the product is a vendor attestation your regulator has already accepted, replacing it means re-running that conversation.
What you escape is the meter. Paying per responder to make people reachable is a pricing model that quietly discourages the exact behaviour paging exists to produce: more people on the rota, more people reachable, thinner escalation steps. Every organisation with a per-seat pager has at some point had the conversation about whether a team really needs a fourth person in the rotation. That conversation should be about sleep, not about $41 a month.
The takeaway
The question is never whether a vendor will fail you. It is whether the day they change their mind is a day you can choose.
Opsgenie is a good product that a reasonable company decided to stop selling. That will happen again, to something else you run, and you will not be consulted. The work is not to predict which one. The work is to know, for each system, how much of it exists only inside the vendor — and to make that number small on the systems where a forced migration would be an emergency rather than a nuisance.
Prices quoted here are vendor list prices with the date we checked them. If a figure has gone stale, tell us and we will correct it — support@conseiltek.com.