Incident Postmortems in India: CERT-In's 6-Hour Rule
Run a blameless postmortem that stands up to the CERT-In directions: an IST timeline on synced clocks, the 6-hour report, 180-day logs and a worked UPI outage.
By Redbison Dev · · 6 min read

India edition. Written for Indian teams: the CERT-In directions of 28 April 2022 under section 70B of the IT Act (6-hour reporting, 180-day logs, NTP sync with NIC or NPL) and a payment outage worked in IST. US edition
For an Indian company, the first deadline in a serious incident is usually not the customer update or the internal review. It is CERT-In's. The directions issued on 28 April 2022 under section 70B(6) of the Information Technology Act, 2000 require service providers, intermediaries, data centres, body corporates and government organisations to report listed cyber incidents to CERT-In within 6 hours of noticing them, to keep logs of all ICT systems for a rolling 180 days, and to synchronise system clocks with the NTP servers of NIC or NPL. A blameless postmortem built on that timeline is both the team's learning document and the evidence that the report went in on time.
This edition shows how, with a UPI payment outage at a Bengaluru fintech built in the Incident Postmortem Generator. It covers the CERT-In directions only. If you are regulated by a sector regulator, or personal data is involved, other reports may also be due, so check with your compliance and legal teams.
What the CERT-In directions require
| Direction | What it says | Why it matters for the postmortem |
|---|---|---|
| (i) Clocks | Connect to the NTP server of the National Informatics Centre (NIC) or National Physical Laboratory (NPL), or to servers traceable to them, for all ICT system clocks | The timeline can be trusted across systems |
| (ii) Reporting | Report the incidents listed in Annexure I within 6 hours of noticing them or being told about them, by email (incident@cert-in.org.in), phone (1800-11-4949) or fax | The timeline must show when the incident was noticed and when it was reported |
| (iii) Point of contact | Designate a point of contact for CERT-In and act on its orders within the time it sets | Name the contact in the response section |
| (iv) Logs | Keep logs of all ICT systems securely for a rolling 180 days, within Indian jurisdiction, and provide them with an incident report or when directed | The postmortem's evidence is still available when CERT-In asks |
Annexure I lists 20 types of incident, including unauthorised access to IT systems or data, malicious code such as ransomware, attacks on servers and network devices, denial of service, attacks on e-commerce and other applications, data breaches, data leaks, attacks affecting digital payment systems, and attacks on cloud systems. The directions took effect 60 days after issue.
CERT-In's FAQs on the directions add three practical points:
- Report what you have. If the incident reporting form cannot be completed within 6 hours, give the information available and send the rest later within reasonable time (Q 30). Data breaches and data leaks are named among the incidents to report within 6 hours.
- Synced clocks, not necessarily IST. Organisations may use a time source other than NIC or NPL as long as it does not deviate from them (Q 40 to Q 42). CERT-In's FAQ Q 40 says there is no need to set system clocks to IST. NTP gives UTC, and the time zone must be recorded with the time so it can be converted accurately; what matters is that every system agrees.
- Which logs. Examples include firewall, IPS, SIEM, web, database, mail, proxy, VPN, SSH and application logs, recording both successful and failed events (Q 37). CERT-In explains that without accurate timestamps it is very hard to recreate the sequence of events across systems (Q 39).
Worked example: a UPI collect outage that became a data leak
A payment aggregator in Bengaluru pushes a configuration change to its UPI collect API. Payments start failing, and the same change switches on a debug endpoint. All times are IST, from servers synced to NIC's NTP service, and the first line of the timeline says so.
In the Incident Postmortem Generator, enter one event per line:
10:02 IST - Collect request failures rise after a config push to the UPI gateway service
10:09 IST - Error-rate alert pages the on-call engineer
10:15 IST - SEV1 declared; merchants informed of payment failures
10:44 IST - Config rolled back; collect requests succeed
11:40 IST - Security engineer notices the same config enabled a debug endpoint returning customer VPAs and mobile numbers; access logs show calls from an unknown external IP (CERT-In 6-hour clock starts)
12:05 IST - Debug endpoint disabled at the gateway
13:20 IST - Permanent fix deployed: debug flags removed from production config
15:10 IST - Incident reported to CERT-In by email with available details and the log extract
The generator splits each line into its time ("10:02 IST") and its event. Enter 7 minutes to detect, 42 to mitigate and 198 to resolve, and the report shows 7 min, 42 min and 3 h 18 min.
Now check the timeline against the directions:
- Two incidents in one. The outage itself was a failed config change. The data leak, noticed at 11:40 IST, is the one that falls under Annexure I ("data leak"), and the 6-hour clock runs from that moment to 17:40 IST. The report at 15:10 IST went in 3 h 30 min after noticing, with the remaining details to follow.
- Logs. The 180-day log retention means the gateway's access logs for the debug endpoint, including the unknown IP's requests, are kept and can be handed to CERT-In. The postmortem should say where they are stored and who can retrieve them.
- Clocks. Because the gateway, the WAF and the SIEM all sync to NIC, the 11:40 IST discovery and the earliest malicious request (found in the logs at 10:21 IST) can be compared directly. Without that, the review could not say how long the endpoint was exposed before anyone noticed.
The trigger was the config push. The root cause is that one config file controlled both production routing and debug features, with no check that debug flags were off in production, and no alert for new endpoints appearing at the gateway. Action items: separate debug flags from production config (solution); fail the deploy pipeline if any debug flag is on in production (prevention); alert when the gateway sees a route not in the API catalogue (detection); a runbook step that asks "does this incident involve data or access?" at declaration, so the CERT-In clock is checked from the start (mitigation). The generator groups them so you can see each category is covered.
Running the review in an Indian team
- Keep it blameless. The question is what let a debug flag reach production, not who flipped it. People who expect punishment leave out exactly the details CERT-In may ask for.
- Make the 6-hour check part of declaring an incident. Put the Annexure I list in the incident runbook. Most outages are not reportable; the ones that are often look like ordinary outages at first, as in the example.
- Write in one time zone. If vendors send logs in UTC, convert them before the review meeting; the Time Zone Converter helps.
- Leave unknowns blank. The generator prints "Not recorded" for an empty duration. That is better than a guess in a document that may go to CERT-In.
- Review within a week, share the draft first, have someone other than the incident commander facilitate, and track action items in your normal sprint tool until closed.
For long causal chains, the Root Cause Analysis tool walks through 5 Whys and fishbone diagrams.
Tools for this job
- Incident Postmortem Generator: timeline, durations, root cause and grouped action items as a Markdown report.
- Root Cause Analysis: 5 Whys and fishbone diagrams.
- Time Zone Converter: putting UTC vendor logs on IST.
For the general method, see the blameless postmortem guide.
Frequently asked questions
Within how many hours must a cyber incident be reported to CERT-In?
Within 6 hours of noticing it or being told about it, for the incident types listed in Annexure I of the CERT-In directions of 28 April 2022. CERT-In's FAQs say you may report the information available within 6 hours and send the rest later.
Is a normal outage reportable to CERT-In?
Not by itself. The directions apply to the listed cyber incidents, such as unauthorised access, malware, denial of service, data breaches and data leaks. An outage from a bad deployment is not on the list, but check whether the same change also exposed data or systems, as in the worked example.
How long must logs be kept under the CERT-In directions?
For a rolling period of 180 days, kept securely within Indian jurisdiction, and provided to CERT-In with an incident report or when directed. CERT-In's FAQs add that copies may also be stored outside India as long as the logs can be produced to CERT-In in reasonable time, and list firewall, SIEM, server, VPN and application logs as examples.
Do we have to log in IST?
No. CERT-In's FAQ Q 40 says there is no need to mandatorily set system clocks to IST: NTP servers give UTC, and the conversion to local time happens on the host. The directions require clocks synchronised to the NTP servers of NIC or NPL, or to a source that conforms to them, and the time zone must be recorded along with the time so it can be converted accurately. Writing the postmortem in IST is fine; state the time zone at the top of the timeline.
Sources and further reading
- Directions under section 70B(6) of the IT Act, 2000, CERT-In, No. 20(3)/2022-CERT-In, 28 April 2022. NTP sync, 6-hour reporting, Annexure I incident list, 180-day logs.
- FAQs on the Cyber Security Directions of 28.04.2022, CERT-In, May 2022. Partial reports within 6 hours, time sources, log types.
- Cyber Security Directions page, CERT-In. The directions, FAQs and the June 2022 extension notice.
- Postmortem Culture: Learning from Failure, Google, Site Reliability Engineering, chapter 15.
- Postmortem Practices for Incident Management, Google, The Site Reliability Workbook.
- Postmortem Process, PagerDuty.
Open the Incident Postmortem Generator
More Documentation & Review
Every tool in the Documentation & Review section, free and in your browser.