
Your first on-call week: a practical survival guide
Your name is on the rotation for the first time. Maybe you came up from the help desk, maybe you joined a small team where everybody takes a turn. Either way, for seven days the pager is yours, and you are probably more nervous than you would admit.
When I led a help desk, the people who handled their first on-call week well were not the ones who knew the most. They were the ones who had prepared the boring things, asked for help early, and wrote things down. This guide walks through one week, day by day, the way a first rotation usually unfolds. Your week will differ in the details, but the shape tends to be the same.
The week before: check the boring things
Most first-week trouble is not technical. It is a VPN token that has expired, a phone that silenced the page, or not knowing who to call. Spend an hour on these before your shift starts:
- Log in to everything you might need, from home, on the laptop you will actually use: VPN, the jump host, the monitoring system, the ticket system, the password vault, the cloud consoles or device management pages. Do it at night if you can, from your own network, because that is where you will be when it matters.
- Ask a colleague to send a test page. Confirm it rings through do-not-disturb, at full volume, with the phone in another room.
- Write down the escalation path: your secondary, your team lead, and the vendor or ISP support numbers with the contract or circuit IDs they will ask for. Keep a copy on paper. If the problem is that the network is down, the wiki might be down too.
- Find the runbooks and read the ones for the alerts that fire most. You do not need to understand everything. You need to know where to look.
- Shadow part of a shift if your team allows it. Watching someone else acknowledge, assess and hand off a page is worth more than any document.
- Plan your life around the week. Keep the laptop charged and near you, avoid long drives through dead zones, and tell the people you live with what the phone noise means.
Monday, 10:00: the handoff
The outgoing person hands over. Ask three questions, even if their notes look complete:
- What is flaky right now, and what does it look like when it misbehaves?
- What changes are scheduled this week, and who is running them?
- Is anything open that might come back?
Then start your log. A plain text file or a running document is fine. Every time something happens, write the time, what you saw, and what you did. It feels silly when nothing is happening. By Thursday you will be glad of it, and on handoff day it writes half your note for you.
Tuesday, 02:47: the first page
The phone goes off and your heart rate jumps. That is normal. Here is the order that works:
- Acknowledge the page. That stops it escalating to the next person while you look, and tells the team someone has it.
- Read the whole alert, slowly. Which host, which check, since when, what threshold. A lot of early confusion comes from reading only the first line.
- Open the runbook for that alert, if there is one, and follow it.
- Work out the scope. Is it one device or many? One site or all of them? A single disk filling up is a different night from a whole branch dropping off the network.
- Stop the bleeding before you look for causes. Fail over, restart the stuck service, free the disk space. The root cause can wait for daylight. Users cannot.
- Say something. Post a short status line where your team expects it: "Looking at the VPN alert for the north office. Next update in 20 minutes." Silence makes everyone else nervous, including your manager.
Set yourself a time limit for being stuck. Fifteen minutes without progress is a sensible default. When you hit it, call your secondary. Escalating is not a failure and nobody experienced will think it is. The engineers people trust most are the ones who escalate early, with a clear summary, rather than late, with an outage that grew while they were proud.
When it is fixed, write it in the log, close or annotate the alert, and go back to sleep. If you cannot sleep, that is normal too. Note it, because it matters on Thursday.
Wednesday: the quiet day
Nothing pages. This is the day to make next week easier for whoever follows you:
- Fix the runbook you used last night. You found the step that was missing or wrong; write it in while you remember. Runbooks an on-call engineer can use at 3 a.m. has a template if yours has none.
- Note every alert that fired without needing action. Those are candidates for tuning, and the people who own the alerts will want the list. Alert fatigue explains why that list matters more than it looks.
- Read two more runbooks for alerts you have not seen yet.
Thursday, 11:20: the alert that is not yours
Packet loss alerts on the WAN link to the warehouse. You check the router, the switch, the firewall. Everything on your side looks clean. The problem is upstream, at the provider.
This is where new on-call engineers lose an hour, because there is nothing to fix and that feels wrong. Your job here is different: get the right people moving and keep everyone informed.
Open a ticket with the provider and give them what they will ask for anyway: the circuit ID, the exact time it started, what you see (loss percentage, latency, which direction), and what you have already ruled out on your side. A traceroute and a graph help. Then post a status line internally so the warehouse knows it is being worked on, and set yourself a reminder to chase the provider if you have heard nothing in an hour.
Friday, 18:40: the big one
A change that was supposed to be routine has gone wrong, and several services are down. More people join the call. It gets noisy.
In an incident with many people, someone needs to coordinate, someone needs to fix, and someone needs to keep notes and talk to the people affected. If your team has an incident lead, let them lead. If nobody is coordinating, it is fine to say "Shall I take notes and send the updates?" That job is valuable, and it is one a first-week engineer can do well. Your Tuesday log habit is exactly the skill it needs.
Do not stay quiet if you know something useful. "The same alert fired Tuesday at 02:47 and cleared after a service restart" is the kind of sentence that shortens an incident.
The weekend: rest is part of the job
On-call does not mean awake. Sleep when you can, eat properly, and keep the laptop close. If you need two hours off for something you cannot move, arrange cover with your secondary in advance and write it down. That is normal practice in any healthy rotation.
If Tuesday's page and Friday's incident have left you short of sleep, say so. Many teams let you hand off a shift or take time back after a heavy night. You can only use that if you tell someone.
Monday again: the handoff note
Your log does most of the work now. A short, consistent note is better than a long story:
On-call handoff, week ending Monday 10:00
Open issues
- Warehouse WAN loss: provider ticket open, last update Sunday 16:00, they suspect a card on their side.
Pages this week
- Tue 02:47 VPN north office down: stuck service, restarted, runbook updated.
- Thu 11:20 Warehouse WAN loss: upstream, see open issues.
- Fri 18:40 Change incident: rolled back 19:55, notes in the incident ticket.
Noisy alerts (no action needed)
- Disk 80% on backup server: fired 6 times, suggest raising threshold or cleaning up.
Changes I made
- Edited runbook for VPN service restart (added the step to clear the session table).
Hand it over in person or on a call if you can, and give the next person the same three answers you asked for on Monday.
After the week
Take a moment once it is over. Write down what surprised you, what you did not know where to find, and which alert woke you for no reason. Bring that list to your team. First-time on-call engineers see the gaps more clearly than anyone, because nothing is familiar yet.
It is also good career material. Being able to describe an incident calmly (what you saw, what you did, when you escalated, what changed afterwards) is exactly what interviewers for NOC and network roles listen for, as the help desk to network engineer article describes. The free Site Reliability Engineering book has a chapter on being on call that is worth reading once you have a week behind you.
The thing to remember: prepare the boring things, escalate early, and write everything down. Those three habits carry you through almost any first week.
next: Uptime and SLA calculator for the numbers, the Rack list for the gear


