When a store’s internet drops or a clinic’s printer server stops, the first report usually comes from a customer. Network monitoring turns that around, and it doesn’t need a big IT team to start.
What to watch first
You don’t need to monitor everything on day one. Start with the things that stop a location from working:
- The internet connection and the router or firewall at each site.
- The core switch and the access points people connect to.
- The servers and services the location depends on, such as a booking system or a file server.
- Your public websites and certificates, which customers see before they ever visit.
Checks that cover most problems
A handful of check types catch most outages:
- Ping shows whether a device answers at all.
- SNMP reads interface traffic, errors, CPU, memory and disk from network equipment and servers.
- HTTP and HTTPS checks confirm a website or web app loads, and warn before a certificate expires.
- TCP port and DNS checks confirm that a service is listening and a name still resolves.
Checking every 5 minutes suits most devices. Save 1-minute checks for the few things where minutes matter.
Check private networks from the inside
Devices inside a location usually aren’t reachable from the internet, and they shouldn’t be. A small collector program at each site can check them from the inside and send results out over HTTPS, so no firewall ports need to be opened. Two collectors at the same site can share the work and cover for each other.
Make alarms worth reading
Monitoring fails when people learn to ignore it. A few habits keep alarms useful:
- Merge repeats so one flapping link doesn’t send fifty messages.
- Let alarms clear themselves when the problem goes away.
- Hold back dependent alarms: if a site’s router is down, you don’t need an alarm for every device behind it.
- Send each alarm to the right person by email, text, Slack, Teams or PagerDuty, with a delay and an escalation step for anything not acknowledged.
- Use maintenance windows for planned work.
Back up configurations
A failed switch is quicker to replace when its configuration is saved. Daily configuration backups, with the differences between versions highlighted, also show who changed what, and when.
Don’t forget the business side
An expired connection to an ad account, a broken booking form or a store whose checkout is down costs money just like a dead router. Watching websites and connected accounts with the same tool means one place to look.
How Metric Sensei helps
Network Monitoring uses collectors at each location and cloud checks from two regions, with discovery, maps, configuration backups, SNMP, traps and syslog, and alarms by email, text, Slack, Teams, PagerDuty or webhook. Every website and connected account in Metric Sensei is watched too, and the AI assistant can read an alarm with its context and suggest the likely cause.