The Essential Utility Belt: Free Tools Every Admin Should Have
Somebody says "the network is slow." You have no monitoring system yet, no budget for one this quarter, and a ticket queue that is not getting shorter while you think about it. I have stood exactly there, with a budget line that read zero and a manager in the doorway. What you do have — or can have, free, in about an hour of installing — is a set of small, sharp tools. They answer the questions behind the complaint. Is it reachable? What does it say about itself? What is really crossing the wire? How fast is the path? Has it always been like this?
This guide is the utility belt. It is not organized by tool name, because nobody troubleshoots by thinking "I shall now use tcpdump." It is organized by the question you are trying to answer, because that is how a problem arrives: as a complaint, not a command. For each question you get the tool that answers it and when to reach for it. You also get the one habit that turns it from something installed into something used. By the end you will know how to build a jump bag that carries all of it. You will understand the honest limit of a utility belt. That is where the rest of this topic takes over.
In this guide
- Why organize tools by question, not by name
- Is it there? Ping, traceroute, and MTR
- What does it say about itself? SNMP tools
- What is actually on the wire? tcpdump and Wireshark
- How fast is the path? Throughput testers
- How has latency behaved over time? SmokePing
- Can I glue it together? Scripting and schedulers
- The jump bag: one place with everything installed
- The belt at a glance
- The honest limit of a utility belt
Why organize tools by question, not by name
Most of us start by collecting tools the way some people collect kitchen gadgets: install everything, use two, forget the rest exist. The fix is to attach each tool to a question, because questions are what you have in the moment of trouble. When a user reports that the file server is "gone," your brain does not need a list of forty utilities. It needs one question — is it there? — and the tool that answers it in five seconds.
Six questions cover most network trouble, and each has one or two free tools that answer it. Learn the six, practice each tool once on a quiet afternoon, and you will have a belt you can draw from under pressure. Everything below runs on Linux, and nearly all of it on Windows too.
Is it there? Ping, traceroute, and MTR
What the question means
"Is it there?" is really three questions folded together. Does the device respond at all? How long does a round trip take? And if it does not respond, where along the path does the trouble start? The oldest tools on the belt answer all three using ICMP (Internet Control Message Protocol), the network's built-in "are you alive?" language. (It has no word for "why".)
The tools
ping sends a small packet and times the reply. A reply means the device is reachable. The time tells you how far away it is in network terms. Missing replies mean something is being dropped. traceroute (spelled tracert on Windows) sends probes with increasing time-to-live values. This is a hop counter that each router lowers by one. The router replies with an error when the counter reaches zero. So each router along the path reveals itself in turn, giving you a hop-by-hop map. MTR combines the two. It runs traceroute continuously and shows per-hop loss and latency as a live table. That is the difference between a snapshot and a movie.
# Is it there, and how far? ping -c 10 10.20.0.5 # Which hops are on the way? traceroute 10.20.0.5 # Where along the way do things get lossy? (Ctrl+C to stop) mtr 10.20.0.5
When to reach for it
First, always. Before you open a single log or admin console, ping the thing. If ping succeeds, the "outage" has just become a "something higher up the stack" problem, which changes your next move entirely. If ping fails, traceroute tells you whether the failure is next door or four routers away. In that case, MTR tells you whether a suspicious hop is truly losing packets or merely slow to answer ICMP. Reading an MTR table without being fooled by such hops is covered in the guide on MTR path testing.
The one habit
Ping from two places. Your workstation is one vantage point; a server in the data center or a machine at the remote site is another. If the target answers one and not the other, the problem is on the path, not the target. In that case, you have just saved an hour staring at a healthy server.
What does it say about itself? SNMP tools
What the question means
A device that answers ping is alive, but alive is a low bar. Is a port down? Is the CPU pinned? Did it reboot ten minutes ago? Most network gear will tell you if you ask in SNMP (Simple Network Management Protocol), the standard question-and-answer language monitoring systems use behind the scenes. The command-line SNMP tools let you ask directly.
The tools
The free Net-SNMP toolkit gives you snmpget for one fact and snmpwalk for everything in a section. Their most useful trick is showing you what a device is willing to tell you. That is exactly what you need when you later configure a monitoring system and wonder why a graph is blank.
# Identity and uptime: did it reboot? snmpwalk -v2c -c readonly 10.20.0.1 system # Every port name and its up/down state snmpwalk -v2c -c readonly 10.20.0.1 IF-MIB::ifDescr snmpwalk -v2c -c readonly 10.20.0.1 IF-MIB::ifOperStatus
When to reach for it
Reach for SNMP tools when a device is reachable but you suspect something inside it. Examples are an uplink that flapped, an interface with errors, a UPS (uninterruptible power supply) on battery, or a switch with suspiciously short uptime. Also reach for them the first time you meet any new device, to learn its vocabulary before you try to monitor it. A full from-scratch walkthrough, from installing the tools to loading the dictionaries that turn numbers into names, is in the snmpwalk tutorial.
The one habit
Save one full walk per device type to a dated text file in a folder you can find. When a device misbehaves, a fresh walk compared against the saved one shows what changed. That could be a port now down, a description someone edited, or a counter that reset. The walk also doubles as documentation you did not have to write.
What is actually on the wire? tcpdump and Wireshark
What the question means
Sometimes every device says it is fine and the application still does not work. At that point you stop asking devices for their opinion and look at the packets themselves. Packet capture records the actual frames crossing an interface. Reading them tells you what the two sides of a conversation really said — including the part where one side stopped replying. I put off learning capture for a long time because it looked like a specialist's tool. It is not; it is the tool for the afternoon when everything else has said "fine".
The tools
tcpdump is the command-line capture tool: small, scriptable, present on nearly every Linux system and most firewalls, and perfect for a server with no graphical desktop. Wireshark is the graphical analyzer. It opens the capture file, decodes hundreds of protocols into readable fields, and lets you filter down to the one exchange that matters. They are a pair: capture with tcpdump where the problem is, analyze with Wireshark where your eyes are.
# On the server: capture only traffic to and from one client, save to a file sudo tcpdump -i eth0 host 10.20.5.77 -w client-problem.pcap
When to reach for it
Reach for capture when the symptom is specific and repeatable but nothing else explains it. Examples are a login that hangs for exactly thirty seconds, a printer that works from one subnet and not another, or an application that "randomly" resets connections. Capture also settles "it must be the network" arguments. A capture showing the server took eight seconds to answer a request the network delivered in a millisecond ends the debate. Reading one — filters, following a stream, spotting retransmissions — is covered in Wireshark fundamentals.
The one habit
Filter at capture time, not just at analysis time. A capture of everything on a busy server interface becomes a huge, noisy file within a minute. Capture one host or one port for one reproduction of the problem, and the file will be small enough to actually read.
Tip: capture files contain real traffic, which may include credentials or personal data. Treat them like any sensitive file. Store them appropriately and delete them when the case closes. Never send a capture to anyone outside without checking what is inside it first.
How fast is the path? Throughput testers
What the question means
Latency and reachability tell you the path is open; they say nothing about how much it can carry. "The backup takes all night" and "the copy to the branch is glacial" are throughput questions. The only honest answer comes from pushing a known amount of data across the path and timing it.
The tools
iperf and its relatives are throughput testers. One copy runs as a server on one end. Another runs as a client on the other. The client reports how many megabits per second the path actually delivered. Because you control both ends, the number is free of application overhead — it is the path's capacity, not the file server's mood.
# On the far end (branch office server): listen iperf -s # On your end: send for 30 seconds and report throughput iperf -c 10.30.0.10 -t 30
Binary names and flags vary a little between generations of the tool. That includes the flag that reverses direction so you need not walk to the other site. So check the help output. The shape of the test is always the same: a listener, a sender, a duration, and a result in bits per second.
When to reach for it
Reach for a throughput tester when someone asks whether a link is "really" the speed the provider sells. Reach for one when a new site connection arrives and you want a baseline before users move in. Also reach for one when a slow-transfer complaint needs splitting into "the path is slow" versus "the application is slow." Test both directions; asymmetric results are common and usually point straight at the cause.
The one habit
Write the result down with the date and time. A single throughput number is nearly useless; the same test at a quiet hour and a busy hour, or this quarter and last, is evidence. When the provider says nothing has changed, you will have numbers that say otherwise.
How has latency behaved over time? SmokePing
What the question means
Every tool so far answers a question about right now. Intermittent problems laugh at right now. The link was fine when you tested it. It was terrible at ten past two, then fine again by the time anyone told you. "How has this behaved over time?" needs a tool that keeps asking while you do something else.
The tool
SmokePing pings a list of targets on a schedule, several probes per round. It draws graphs where the spread of results appears as smoke-like shading around the median (the middle value of each round). Loss is shown as color changes. A glance at a week of graphs tells you whether latency is steady and whether it spikes every afternoon. It also tells you whether last night's "the VPN was awful" report matches anything measurable.
When to reach for it
Set it up before you need it, pointed at your gateway, your provider's first hop, each remote site, and one or two external references. The value is entirely in the history; a grapher installed during an incident has nothing to show. Setup, probe settings, and reading the smoke are covered in the guide on SmokePing latency graphing.
The one habit
Look at the graphs on a quiet day. If you only open them during outages you will not know what normal looks like, and normal is the whole point. Two minutes each Monday morning is enough to learn the personality of each link. I once spent a week hunting a four o'clock latency spike that, according to the graphs, had been there since the day I installed them.
Can I glue it together? Scripting and schedulers
What the question means
Every tool above is a hand tool. The final question is whether you can make one run by itself. For example, ping twenty hosts every five minutes and record which failed. Or walk a switch nightly and diff the result (compare it line by line against last night's). That is scripting plus a scheduler, the bridge between a utility belt and actual monitoring.
The tools
On Windows, PowerShell plus Task Scheduler. On Linux, a shell script or Python plus cron. Both ship with everything you need. A script does not have to be clever. The useful ones are boring: loop over a list, run a tool, append a line to a log, and email if something changed. Here is the entire idea in a few lines of PowerShell:
# reach-check.ps1 - ping each host in hosts.txt, log the ones that fail
$stamp = Get-Date -Format "yyyy-MM-dd HH:mm"
foreach ($h in Get-Content C:\belt\hosts.txt) {
if (-not (Test-Connection -ComputerName $h -Count 2 -Quiet)) {
Add-Content C:\belt\reach.log "$stamp DOWN $h"
}
}
And the cron line that runs the shell-script equivalent every five minutes on Linux:
*/5 * * * * /opt/belt/reach-check.sh >> /var/log/reach.log 2>&1
When to reach for it
Reach for a script the second time you do something by hand. The first time is exploration; the second is a pattern. The third without a script is a choice you will regret at two in the morning. Scripts excel at "did anything change?" jobs: nightly walks, config backups, port status snapshots.
The one habit
Put every script in a backed-up folder — ideally in version control. That is a system that keeps every past copy of a file and records who changed it. Put a comment at the top of every script saying what it does and where it runs. A script nobody can find runs until it silently stops, and nobody notices until the thing it watched is on fire. The guide on the hidden costs of cheap explains where that road leads; this habit is the cheap way off it.
The jump bag: one place with everything installed
A utility belt is only useful if it is on you. The jump bag is a laptop or a small virtual machine with every tool above installed, tested, and documented before you need it. Firefighters do not assemble their equipment at the fire; you should not be running package installers while a site is down. The checklist:
- One machine, both worlds. A laptop with a Linux environment alongside Windows, or a small Linux VM you can reach from anywhere. That doubles as a permanent second vantage point for ping.
- All six questions covered. ping, traceroute, MTR, the Net-SNMP tools, tcpdump, Wireshark, a throughput tester, and a scripting environment. SmokePing lives on a server, not in the bag; bookmark its graphs.
- Credentials and addresses, safely. Read-only SNMP community strings, the IP list of core devices, and logins for your second vantage point, in your password manager. They do not belong in a text file on the desktop.
- A one-page cheat sheet. The commands in this guide with your own addresses filled in. Under pressure you forget the flag you use twice a year.
- The saved baselines. Reference SNMP walks, throughput results, and a recent MTR run to each site, so "what changed?" has something to compare against.
- A quarterly test. Boot it, run one command per question, fix anything that broke. Jump bags decay.
Tollgate Foods had one IT person. Her whole belt lived on one laptop. It held the scripts, the saved walks, the cheat sheet, and the throughput baselines. It also held a reach-check log that went back further than the ticket system did. The laptop was stolen from a car on a Tuesday. Nothing on it had been backed up, because the backups were one of the things the scripts did. She rebuilt the belt over a month, partly from memory and partly from two commands she had once pasted into a ticket. The rebuilt belt lives in version control with a copy on a server. She now describes that arrangement to every new starter as "the bag, and the other bag".
If you are the only IT person in the building, the jump bag matters doubly. The guide on a solo monitoring setup goes further into arranging things so one person can cope.
The belt at a glance
| Tool | Question it answers | Reach for it when | The habit |
|---|---|---|---|
| ping | Is it there, and how far? | First step of every "it's down" report | Test from two vantage points |
| traceroute / tracert | What path does traffic take? | Ping fails or latency is odd | Save a known-good trace per site |
| MTR | Which hop loses packets? | Intermittent loss on a path | Run long enough to see the pattern |
| snmpwalk / snmpget | What does the device say about itself? | Reachable but misbehaving; any new device | Keep a saved walk to diff against |
| tcpdump | What is on the wire, right here? | Devices say fine, application disagrees | Filter to one host or port at capture time |
| Wireshark | What did the conversation actually say? | Reading a capture; settling "it's the network" | Learn three filters and follow-stream |
| iperf-style tester | How much can the path carry? | Slow transfers, new links, provider disputes | Record date and direction with each result |
| SmokePing | How has latency behaved over time? | Before the intermittent complaint arrives | Read the graphs on quiet days |
| PowerShell / Python + scheduler | Can I make it run by itself? | The second time you do anything by hand | Version control and a header comment |
The honest limit of a utility belt
Everything on this belt shares one property: you have to pick it up. Ping does not run unless you type it. A capture does not start unless you were already suspicious. The belt is reactive — magnificent once you know something is wrong, and silent until then.
That silence is the gap. A user notices the outage before you do. A link runs at half capacity for a month because nobody ran the test. A switch reboots nightly for a week before someone walks its uptime. The tools were all there; nobody was asking at three in the morning, because people sleep and tools do not volunteer.
Closing that gap is what a monitoring stack does: it asks the belt's questions continuously, keeps the answers, and speaks up when they change. You can build one from free components. The next guide, free stacks that punch above their weight, shows how the pieces fit. Or you can buy a stack. Either way, the belt stays; every alert the stack raises sends you straight back to ping, a walk, or a capture to find out why. The belt is where troubleshooting happens; the stack makes sure it starts on time.
Where to go next
Build the jump bag this week, not someday. Install the tools, run each one against something real, and save a baseline walk and a trace. Write the cheat sheet with your own addresses on it. On the next quiet afternoon install SmokePing and point it at your sites, so the history exists before the next intermittent complaint arrives. None of this needs the budget line to change. From there the rest of the Monitoring on a Budget: Free and Low-Cost Stacks topic picks up. It covers assembling a free stack, where a little money actually helps, and the hidden costs of cheap. It also covers a staged roadmap from the belt you now own to monitoring you can trust. The next time someone stands in the doorway and says the network is slow, you will have a question ready, and a tool for it.
Put this into practice with SysPeep Monitor
Every tool on this belt answers a question you have to remember to ask. SysPeep Monitor asks them for you from a single Windows machine. It discovers devices, polls them over SNMP and other collectors, and alerts when the answers change. It is a commercial product with a free trial rather than a free tool. So try it alongside your belt and decide whether the time it saves earns its place.
Download Free TrialLearn About SysPeep