The Monitoring Handover Pack: A Checklist You Could Hand Over Today
If you had to hand the monitoring to someone else this afternoon, what would you give them? Not what you would explain over a coffee; what you would put in their hands and walk away from. For most administrators the honest answer is a password and an apology. The monitoring works, and the reasons are in your head. The pieces of paper that exist are scattered across a wiki page, a text file on your desktop, and a spreadsheet from three years ago. They are also in the description fields of about a third of the alert rules.
The handover pack is the answer to that question, assembled in advance. It is the small set of documents that lets a competent administrator run the monitoring from day one. It is kept in one place and reviewed on a date. It is tested once a year by the only test that counts: someone else running the monitoring while you are genuinely unavailable. This guide describes what goes in it, how to test it, and where it lives. It gives you a checklist with a "where it is" and a "last checked" column for every item.
The earlier guides in this pillar produce the pieces. Handover-ready monitoring writes the README and the map; extracting tribal knowledge fills in the reasons; onboarding the next admin proves they work. This guide is where the pieces are gathered, named, and kept.
In this guide
- The pack, defined
- The eleven items
- The annual test
- The pack as a living thing
- Where the pack lives
- The handover pack checklist
- Checking the pack's freshness
- An honest note about being replaceable
- The checklist
- Where to go next
The pack, defined
The handover pack is everything the next person needs to run the monitoring, in one folder. Its cover sheet says what is in it and when it was last checked. It is not a manual for the product; the vendor wrote one. It is not a description of the network; that is the network documentation, which the pack sits beside. It is the monitoring-shaped set of facts that only your organisation could have written. They are about your checks, your thresholds, your alert routing, and your reasons.
"One folder" is a real requirement, not a figure of speech. A pack whose parts live in four places is a scavenger hunt. The newcomer does not know which parts exist, so they cannot know when they have found them all. One folder, one cover sheet listing the contents, and a rule that anything not in the folder is not in the pack. If the runbooks live elsewhere, the pack holds an index that points at them, and the index is the pack's part.
The test of a pack is the one from the bus factor of your monitoring. Could a competent colleague, tomorrow, with no call to you, add a device, change a threshold, and learn why an alert exists? Could they restore the server and silence a false alarm safely? A pack that answers all five is complete. A pack that answers four is a good pack with a hole in it, and the hole is where the incident will be.
The eleven items
The pack has eleven items. Most are one page; two are longer. Together they are shorter than the product manual and worth more, because the manual describes every monitoring system and the pack describes yours.
- 1. The README. The cover document: what the system is, who owns it, the ten sections from handover-ready monitoring, and the change log. Everything else in the pack is referenced from here.
- 2. The map. One page showing the monitoring's own parts: server, probes, database, backup target, mail relay, alert paths, credential store. Real host names, not placeholders.
- 3. The service-to-checks table. One row per service the business would name, the checks that watch it, one sentence on why it matters, and who cares.
- 4. The threshold reasons. Every non-default threshold with its value, the default it replaced, and the "why this number" sentence.
- 5. The credential-store pointer. Which store, which folder, which entry for each credential the monitoring uses, and which group has access. No secrets in the pack, ever.
- 6. The alert routing table. For every alert group, who is notified first, who after fifteen minutes, who after an hour, and by what means. It must match the tool.
- 7. The backup and restore procedure. Where the backup goes, when, what it contains, the step-by-step restore, and the date of the last successful restore test.
- 8. The maintenance calendar. Every recurring maintenance window, what it suppresses, and the job or event it exists for.
- 9. The known-issues list. The check that is flaky, the device that lies about its uptime, the thing you always meant to fix, each with a workaround.
- 10. The contacts list. Everyone outside the team who is told about, or acts on, a monitoring alert, by name, role, and how to reach them. The printer contact. The person at the ISP who answers.
- 11. The runbooks index. A list of every runbook that a monitoring alert points to, where it lives, and its last review date. Runbook anatomy describes what a good one contains; the pack only needs to know where they are.
Items one to seven are the ones the bus-factor test needs. Items eight to eleven are the ones the second month needs, when the newcomer has stopped surviving and started maintaining. If time is short, do the first seven and put "items eight to eleven: not yet written" on the cover sheet. A pack that is honest about its gaps is still a pack. (A pack that hides them is a brochure.)
The annual test
The pack is tested once a year by the only method that produces a true result. Someone else runs the monitoring for a week while you are genuinely unavailable. Not "available if it is urgent". Not "checking email on the beach". Unavailable, with the phone off, and the colleague told in advance that this is the case. Surviving vacations covers the wider problem of getting a solo administrator out of the building. The handover pack is the part of that problem that concerns the monitoring. The annual test is the vacation with a purpose.
The measurement is the questions. Every question the colleague wanted to ask you and could not is a gap in the pack. The colleague's job for the week, besides running the monitoring, is to write each one down as it occurs. "Where is the SNMP string for the branch switches?" is a gap in item five. "Why does the file server alert at 91?" is a gap in item four. "Who do I tell when the label printers stop?" is a gap in item ten. When you return, the list of questions is the to-do list, and it is usually shorter than you feared and longer than you hoped.
The week has to be a real week, because the gaps are in the things that happen weekly. A two-day test finds the daily gaps. A full week catches the Sunday maintenance window, the Monday morning backlog, and the Thursday report. It also catches the one alert that only fires when the warehouse does its stock count. Choose a week with something scheduled in it, if you can. A quiet week tests the pack against nothing.
If the colleague had to call you anyway, that is not a failed test; it is a found gap with an urgency rating attached. Write it down with the others and fix it first. If the colleague did not have to call you at all, run the test again next year with a different colleague. The second one will find the questions the first one happened to know the answers to.
The pack as a living thing
The pack stays true by two mechanisms. One is a review date on the cover. The other is a rule that every change to the monitoring updates the matching item. The review date is a calendar appointment, quarterly, in which you read the pack against the console and correct every line that disagrees. The rule is the one from handover-ready monitoring: no change without a line, before the console is closed. Between them, the pack is never more than one change behind the tool and never more than a quarter behind reality.
The cover sheet is what makes the review date work, because it puts the date where the newcomer will see it first. A newcomer who opens a pack with a review date in the past will, correctly, trust nothing in it. The pack will have failed at the one moment it existed for. A pack with a recent date on the cover is read as a set of facts. The date is a promise that someone checked, and the quarterly review is keeping it.
Update items in the pack when the events that touch them happen, not on the review date. A restore test updates item seven the same afternoon. A new maintenance window updates item eight before the window is saved. A contact who leaves the ISP updates item ten when you find out. The review date catches what the rule missed; it is not the mechanism, only the safety net.
Where the pack lives
The pack lives with the network documentation, on a share that is backed up and that the on-call group can read. The share does not depend on the monitoring server or on your laptop to be reachable. Those four conditions rule out the most common locations. A folder on the monitoring server is unreadable during the one incident where the pack matters most. A folder on your laptop leaves with you. A wiki that requires a login only you have is your laptop with extra steps.
"With the network documentation" is deliberate, because the pack's reader will need both, usually on the same day. The network documentation pillar covers what that documentation is and how it survives. The monitoring pack is a subfolder of it, with its own cover sheet, and the network documentation's own index points at it. The monitoring should never be the one system whose documentation is somewhere else.
Keep a copy somewhere that survives the share being down. A printed copy in the server room is not old-fashioned. It is the only version that works when the file server, the network, and the monitoring have all failed at once. That is precisely the day the pack was written for. Print it on the review date and put it in the folder marked with the date. Recycle the old one.
Tip: put the credential store's own recovery procedure in the pack as well, as a pointer. A pack that says "all secrets are in the store" is only useful to someone who can open the store. And the person who can open it after you have gone is the person named in the store's recovery procedure. If that person is also you, that is the first gap to fix.
The handover pack checklist
The checklist is the cover sheet. Every item in the pack has a row, with what it is, where it is, when it was last checked, and whether it passed. Copy it, fill in the "where it is" column with real paths, and update the "last checked" column at every quarterly review. An item with a blank "where it is" does not exist yet, and the checklist says so honestly.
| Item | What it is | Where it is | Last checked | OK? |
|---|---|---|---|---|
| 1. README | Cover document, ten sections, change log | Path or link | Date, initials | Yes / No |
| 2. Map | One-page diagram of the monitoring's parts with real names | Path or link | Date, initials | Yes / No |
| 3. Service-to-checks table | Service, checks, why it matters, who cares | Path or link | Date, initials | Yes / No |
| 4. Threshold reasons | Every non-default threshold with its "why this number" | Path or link | Date, initials | Yes / No |
| 5. Credential-store pointer | Store, folder, entry names, access group; no secrets | Path or link | Date, initials | Yes / No |
| 6. Alert routing table | Per alert group: first notified, escalations, means | Path or link | Date, initials | Yes / No |
| 7. Backup and restore procedure | Backup target, schedule, contents, restore steps, last test date | Path or link | Date, initials | Yes / No |
| 8. Maintenance calendar | Every recurring window, what it suppresses, and why | Path or link | Date, initials | Yes / No |
| 9. Known-issues list | Flaky checks, odd devices, things meant to be fixed, workarounds | Path or link | Date, initials | Yes / No |
| 10. Contacts list | Everyone outside the team an alert reaches, by name, role, and means | Path or link | Date, initials | Yes / No |
| 11. Runbooks index | Every runbook an alert points to, its location, and its review date | Path or link | Date, initials | Yes / No |
| Cover | This sheet, with the pack's review date and the date of the last annual test | Top of the pack folder | Date, initials | Yes / No |
The "OK?" column is a judgement, not a tick for existing. An item is OK when it was checked against the live console and agreed with it. A routing table that exists but names a person who left is "No", and the review is where it turns back into "Yes".
Checking the pack's freshness
The pack's files carry their own evidence of neglect, because a file's last-modified date is a fact the author cannot improve by good intentions. A short command run on the pack folder lists every file with the date it last changed. It flags any that have not changed in a quarter. Treat it as the shape of the check; the path and the threshold are yours to set.
# Illustrative: list the pack's files by age and flag anything untouched for 90 days
$pack = '\\docs01\network\monitoring-pack'
Get-ChildItem -Path $pack -File -Recurse |
Select-Object Name, LastWriteTime,
@{ Name = 'AgeDays'; Expression = { (New-TimeSpan -Start $_.LastWriteTime).Days } } |
Sort-Object AgeDays -Descending |
Where-Object { $_.AgeDays -gt 90 }
A file that has not changed in ninety days is not necessarily wrong; the contacts list may be stable for a year. It is a file to read at the next review rather than skim, because it has had the longest chance to drift. If the routing table or the threshold reasons appear in the list, be suspicious. Those two change whenever the monitoring does. A monitoring system that has not changed in a quarter is either very stable or very neglected.
At Acme the monitoring administrator left in a fortnight with a handover consisting of a password and a wave. Her successor inherited four hundred checks, ninety alert rules, and no way of knowing which of them were true. For six weeks he treated every alert as a rumour. Each one was checked by hand against the device before anyone acted on it. Nobody could say whether a threshold was meaningful or a leftover. The first alert he acted on without checking was in week seven. The review recorded that the monitoring had "remained operational throughout", which was true in the sense that the lights had stayed on.
An honest note about being replaceable
Writing the pack feels, for about an afternoon, like writing yourself out of a job. It is worth saying plainly that it is the opposite. A pack does not make you replaceable; it makes you promotable. The administrator who cannot be replaced also cannot be moved, cannot take the project, and cannot take the holiday. They cannot take the next role, either. That is because the current one has no one to hand to. The pack is what lets you leave the desk, and leaving the desk is how careers move.
The mechanism is trust, and it runs through evidence. Suppose a manager watches a colleague run the monitoring for a week from your pack, without calling you. That manager has seen something no appraisal form captures: that you built a system other people can run. That is a different and rarer skill than building a system only you can run. And it is the one that gets offered the next thing. Monitoring skills and careers covers the longer road. The pack is the first mile of it. It is the mile most people never walk because it looks like it goes the wrong way.
There is also a plainer reason. One day you will leave, and the person who comes after you will open a folder. What they find there is the last thing you will ever say to them. Make it a pack.
The checklist
Use this to decide whether the pack could be handed over today. It could when every line is a yes.
- All eleven items exist, or the cover sheet says honestly which ones do not.
- Every item is in one folder, or is pointed to from the folder's index.
- The cover sheet has a review date, and it is in the future.
- Every item's "last checked" date is within the last quarter.
- The pack contains no secret, and the credential-store pointer names a store at least two people can open.
- The alert routing table matches the tool and names no one who has left.
- The restore procedure has a last-test date within the last year.
- The pack lives with the network documentation, on a backed-up share the on-call group can read, and not on your laptop.
- A printed copy exists in the server room, dated.
- The annual test has been run within the last year, and every question from it has been closed or added to the known-issues list.
- The freshness check has been run and every file it flagged has been read at a review.
- Someone other than you knows where the pack is and has opened it.
Where to go next
The handover pack is eleven items in one folder with a date on the cover. It is tested once a year by a week of your absence. It is kept alive by the rule that every change gets a line. It is the physical form of everything this pillar has argued. The monitoring system, the thing that watches everything else, deserves to be something more than one person can run. Build the first seven items this month. Test it when you next take a proper week off. Fix what the test finds. Then the monitoring survives you, which was the point.
The pieces come from the rest of this pillar. There is the bus factor test that shows the gaps. There are the README and map, and the interview that fills in the reasons. There is the week that proves it works. The habits behind all of them are the ones monitoring fundamentals that last describes. All of it belongs to Making Monitoring Survive You: Documentation and Bus Factor.
Put this into practice with SysPeep Monitor
A handover pack is shorter when the monitoring has fewer parts to describe. SysPeep Monitor runs as a central Windows service with optional remote probes. It keeps devices, dependencies, and alert rules together in one web console. It stores its data in SQLite, MySQL/MariaDB, or SQL Server. So item seven of the pack is a database you can back up, copy, and restore. For exporting rules or checks to include in the pack, check the product documentation. The free trial is at /download/.
Download Free TrialLearn About SysPeep