Skip to content

Network services

DHCP, DNS, NTP, WireGuard, routing, tunnels and PPP, MLAG, logging, NetFlow and discovery, viewed across the fleet rather than one device at a time. Network Services → Overview is a table of which devices run what: search it, sort by any service, and click a row to open the device.

The Overview and the NetFlow fleet view show what was last read from each device and when, refreshing anything older than ten minutes in the background; Refresh reads every device again. On a large fleet the page fills in over a few seconds rather than contacting every device at once (see Scaling).

DHCP, DNS and NTP show one device at a time, chosen at the top of the page. When the selected site or group has only one device, it is chosen for you (from @ronchan44's #464), and so is the router for PPP Users and Sessions when there's only one.

Most of it writes to the device (Routing, Tunnels & PPP and MLAG only read). Changes go through Change Guard where they could affect reachability.

DHCP

Servers, address pools and leases, per device. A lease list shows the first 100; Show all lists the rest.

  • IPv4 Static Leases lists a device's reservations: add one with Static, remove one with its delete button. Dynamic leases are on the Clients page, which knows far more about each client.
  • Pools and servers can be created, edited and removed.
  • A static IPv6 binding is matched by the client's DUID, not its MAC address, and assigns a prefix. Before 0.24.51 the form asked for a MAC address, which RouterOS refused.

A DHCP server bound to an interface that later changes VLAN is a common way to hand out addresses on the wrong segment; the lease list is the quickest way to notice.

DHCP Pool Usage, at the top of the page (and on Wireless → Overview), shows how full every IPv4 address pool across the fleet is, read on each slow poll (#156): yellow from 80%, red from 95%. A pool that runs out is a quiet failure, since clients connect but get no address, so a nearly full pool is also listed on the dashboard. Widen the range or shorten the lease time.

DNS

The resolver configuration on each device, plus static entries.

  • Static DNS entries are managed per device, including regexp entries. The record types are RouterOS's own: A, AAAA, CNAME, MX (with a preference), NS, TXT, SRV (with port, priority and weight) and FWD (forward that name to another DNS server). RouterOS has no static PTR record. Before 0.24.51 MX, NS and SRV targets went into the wrong field and the record wasn't created.
  • Flush cache clears the resolver cache on a device, which is worth doing after changing an upstream server or a static entry.
  • Upstream DNS servers are edited one entry at a time (#207): type an address and press Enter, or paste a list; × removes one. Only IP addresses are accepted. A device's own DNS servers, under its Config tab, work the same way.
  • DNS over HTTPS (DoH) (#468, 1.1): give the router a DoH server URL, such as https://cloudflare-dns.com/dns-query or https://dns.google/dns-query, and it sends its lookups there over HTTPS instead of to the upstream servers. Leave it empty for plain DNS. The router still has to look up the DoH server's own name, so keep an upstream server (or a static A record for that name); the page warns when there's neither. Verify the server's certificate is recommended: recent RouterOS checks it against its built-in trusted roots, and the page says when a router has none, in which case the server's CA certificate has to be imported first or every lookup fails.

NTP

Client and server configuration. Worth more attention than it usually gets: a device with the wrong clock produces logs that cannot be correlated, certificates that appear invalid — see Alerting — and scheduled work that fires at the wrong time.

NTP servers are edited one entry at a time, like the DNS servers; addresses and host names (pool.ntp.org) are both accepted.

Configuration templates can push the same NTP servers to many devices at once.

On RouterOS 6 the NTP server comes with the optional ntp package, so a stock device shows the client only. Its client takes up to two IP addresses plus any number of names, and its mode follows from the servers: unicast when some are set, broadcast when none are. Before 0.24.51 saving NTP settings on such a device failed with an error 500.

WireGuard

Interfaces and peers per device: create, edit, enable, disable and remove. Peer public keys and allowed addresses are shown so a tunnel can be checked against the far end without leaving the platform.

Every device in the site that has WireGuard gets its own card. Devices without it are listed underneath; Show / add interface opens a card for one so an interface can be added.

A peer's persistent keepalive is in seconds; leave it empty to turn keepalive off.

Client config and QR code

View (the eye) on a peer shows the client's .conf and a QR code to scan into the WireGuard app, when the router keeps the client's private key for that peer. Peers made with WinBox's WireGuard wizard do, and so do peers made here with Generate a client key pair now (#444):

  • The key pair is made in your browser. The public key goes into the peer, and the private key is saved on the router with the client's address, DNS and endpoint, as WinBox does. Anyone who can read the router's configuration can read it.
  • Client endpoint host is where the client reaches this router. It starts as the management address, which is often not the address the client should use, so check it. The router's listen port is added unless you give one.
  • Client allowed IPs is what the client sends through the tunnel. Narrowing it from the full tunnel needs RouterOS 7.21 or later.
  • Turning it on when editing a peer replaces the peer's key, so the client using the old one needs the new config.

A peer whose private key isn't on the router has no config to show; View says so. Operators and admins see View; read-only accounts never receive private keys.

Routing

Network Services → Routing lists every BGP peer and OSPF neighbour across the fleet, or the selected site (#279): the device, the peer's address, AS or router ID, whether it's up, how long for, and when it last changed. Down peers are listed first. They're read with each device's slow poll, so a change shows up within that interval. Configure BGP and OSPF on a device's Routing tab, and turn on the Routing peer down alert (Alerting) to be told when a session drops.

  • BGP: one row per configured connection (a peer on RouterOS 6). It's up when its session is established; a connection with no session is idle, one still trying is connecting, and a disabled one is disabled. Adding a BGP connection on a device's Routing tab asks for the name, remote address, remote AS, role and local AS. On RouterOS 7.20 and later a connection belongs to a BGP instance, which carries the local AS: the device's instance is used (you choose when it has several), and on a device with none, one called default is created with the local AS you enter. On 7.0 to 7.19 the local AS goes on the connection itself, and on RouterOS 6 a peer is added. Before 0.59.1 adding a connection failed on 7.20 and later with missing =instance=.

  • OSPF: one row per neighbour. RouterOS forgets a neighbour once it stops hearing it, so the manager keeps the last row and marks it down. A neighbour gone for a week is dropped from the list. 2-Way counts as up: on a LAN, two routers that are neither DR nor BDR stop there by design.

Tunnels and PPP

Network Services → Tunnels & PPP lists, across the fleet or the selected site (#310), every IPsec peer, PPPoE server, PPPoE client, active PPP session, and EoIP, GRE and IPIP tunnel. It shows each one's local and remote address, its state, how long it's been up and when it last changed. Problems are listed first. Everything is read with each device's slow poll, in one batched request, so a change shows within that interval; Refresh asks for a fresh poll. Apart from PPP (below), set these up on the device.

Type One row per Up when
IPsec configured peer, including the dynamic ones an EoIP or GRE tunnel with an IPsec secret creates a peer that matches it is established. A passive peer with nobody connected is waiting, not down
PPPoE server server (service name on an interface) it has sessions; with none it's waiting
PPP session active session in /ppp active (PPPoE, L2TP, SSTP, OpenVPN...) always: an ended session is removed
PPPoE client client interface it's connected
EoIP, GRE, IPIP tunnel interface it's running

An active IPsec peer is matched to the configured peer whose address is the most specific match, as RouterOS does; a peer that takes many clients (0.0.0.0/0) shows how many are connected. RouterOS doesn't say which PPPoE server a session is on, so with more than one server the count covers all of them. Secrets (the IPsec secret on an EoIP or GRE tunnel) are never stored or shown, and PPP passwords aren't read at all. Up to 5,000 sessions are kept per device, and the page shows the first 500 that match the filters.

PPP users and sessions

PPP Users and Sessions, at the top of the page, reads one router live: its PPPoE servers, who is connected, and its PPP users (/ppp secret). Choose the router, or press manage on a PPPoE server or PPP session row. Operators and admins can:

  • Disconnect a session. Its client usually dials straight back in; turn the user off first to keep it out. A session whose address is a managed device asks again, since ending it cuts the manager off from that device until it reconnects.
  • Turn a user off or on. Off stops it logging in; a session it already has stays up until you disconnect it. Each user shows whether it is connected, its last caller and when it last logged out.
  • Turn a PPPoE server off or on. Off ends every PPPoE session on the router, so it asks first.

User and server changes run under Change Guard and can be reviewed first; every action is in the audit log. Passwords aren't read.

MLAG

Network Services → MLAG shows each MLAG pair (#376), and is in the sidebar only once a device in view has a bridge with an MLAG peer port. For each pair: both peers' state, role, priority and settings, the peer link and how many of its members are up, and the MLAG bonds side by side by mlag-id, with the downstream device each one's LACP partner is. The two peers are found from neighbour discovery on the peer link, and the primary is shown on the left. It's read with the slow poll, and only on devices with MLAG.

Between the two peers, the peer link has a block of its own, since it's the first thing to check when MLAG goes wrong:

  • Members side by side: each physical port of the peer link on both switches, up or down (and "not active" when the link is up but LACP hasn't taken it), with the speed, RouterOS's link-downs count and the errors in the last 24 hours. The two ends of each cable are paired from neighbour discovery, or by name when discovery doesn't say.
  • VLANs carried, with VLAN 1 called out, since host sync needs it, and any VLAN that's on an MLAG bond but missing from the peer link.
  • LACP partner: each side's partner should be the other peer. One that's something else means the peer link is cabled to another device, which is a critical finding.
  • Last flap: the last time a member went down that the manager saw (it keeps 7 days), marked (reboot) when one of the peers restarted around then, so a real flap stands out from a maintenance window.

A pair is critical when it has a critical finding, warning when it has only warnings, and healthy otherwise. A bond down on both peers that has never been seen up, for example one set up for a switch that isn't cabled yet, shows as grey unused and isn't a problem. A bond that has been up and is now down on both peers stays red, with a warning.

It checks, worst first:

Check Severity
A peer isn't connected critical
Both peers primary (split brain), or both secondary critical
The peer link is down / some of its members aren't active critical / warning
One mlag-id with different LACP partners on the two peers (cabled to two devices) critical
The peer link's LACP partner isn't the other peer critical
An mlag-id up on one peer only (single-homed) warning
An mlag-id that was up, now down on both peers warning
A VLAN on an MLAG bond that isn't tagged on the peer link, VLAN 1 included. Hosts in it can't sync between the peers: the "peer failed to sync" storm warning
mlag-heartbeat, mlag-init-delay, mlag-lacp-fallback or the RouterOS version differ warning
State or role changed in the last 24 hours (the first value the manager reads doesn't count) warning
Heartbeat timeouts in the last hour (critical over 30), host sync failures in the last hour, peer link down in the last 24 hours, from the devices' logs warning

The MLAG degraded alert (Alerting) reports critical findings. It's on by default, and does nothing on a fleet without MLAG.

Netinstall

Network Services → Netinstall reinstalls RouterOS on a device from a router running MikroTik's netinstall package (#310), instead of from a PC with the Netinstall tool. The package needs RouterOS 7.24 or later, and isn't made for SMIPS devices. The device being installed must be on the same layer 2 segment as that router, and etherboot to it. An install reformats that device: what's kept depends on the entry (below).

Because of that, every change here (installing or removing the package, entries, the cache, starting an install) needs an admin for the router's site: a fleet admin, or a site admin there. Operators and viewers see the page read-only.

The page lists the devices in view in three groups:

  • Netinstall servers: devices with the package installed. Each one is read live: its entries, the devices that have etherbooted to it, and its cache.
  • Can run Netinstall: Install package installs it. The device reboots to do it, and shows as a server once it's back and polled. Installing the package changes nothing else: the device doesn't answer etherboot until an entry is added. Remove package on a server takes it off again, also with a reboot. Either is refused while another package change is waiting on the device, or when its flash is too full. When the device's last update check found a newer RouterOS, RouterOS may install that too while it applies the change (a test switch went from 7.25 to 7.25.1 this way), so the manager asks first. Tick Also upgrade RouterOS to go ahead, or upgrade it from Firmware first.
  • Can't run Netinstall: older than 7.24, SMIPS, or not polled yet.

Entries

An entry makes the server answer etherboot on one interface. Every entry has to name the MAC address of the device it installs, or have Wait on, so each install is started from this page. Without either, any device on that segment that etherboots would be reformatted, so the manager refuses to save or enable one, and flags one made outside it in red. A new entry has Wait on and FlashFig off.

Setting What it does
MAC address Install only this device. RouterOS can't clear it later: delete the entry and add a new one
Address range The addresses handed out while installing, at most 256
RouterOS version Empty installs the latest in the server's update channel
Extra packages Installed with RouterOS, which is always installed
After installing Reboot, shut down, or do nothing
Keep the old configuration Puts the device's configuration and users back after the install (files and containers aren't kept). Off with Apply the default configuration for a clean device
Install each device only once Ignore a device again after it's been installed, until the server restarts
Wait Hold each device that etherboots until its install is started here
Answer etherboot / FlashFig What the entry responds to

Entries are written under Change Guard, and Review changes shows the commands first. Script and mode files aren't offered yet.

Installing a device

A device that etherboots to a server is listed under Devices seen, with its model, architecture and progress. The list refreshes every five seconds while it has devices in it. When the MAC address is one a managed device has, the device is named. Reinstall… starts the install. You can choose a version and packages, and it asks you to type the device's MAC address back first.

Cache

Install files download from MikroTik when an install needs them. Download files fetches them first, so installs don't wait and work without internet access. The header shows where the cache is. On a device with a flash folder (a CRS309, say) the root folder is RAM, so files there use RAM and are gone after a reboot. On other devices they use flash. Cache folder points it at a disk instead, a USB drive or a RAM disk (/disk add type=tmpfs).

Flash and RAM are never filled. Before an entry is enabled, files are downloaded or an install is started, the manager estimates the download (about 20 MB for RouterOS, plus the extra packages) and refuses when it wouldn't fit. It keeps 5% of flash free, and at least 32 MB of RAM.

Logging (syslog)

Where each device sends its logs — remote actions and the rules that select which topics go to them. Rules can be toggled without deleting them.

On current RouterOS 7, a remote action with a facility and severity is sent in syslog format (remote-log-format=syslog), and BSD syslog picks the BSD timestamp style over ISO 8601. Older RouterOS takes the same settings in its own form. The built-in actions (memory, disk, echo, remote) can be edited but not renamed or changed to another kind. Before 0.24.51 adding an action failed on every device.

This is separate from the events the platform collects itself. The platform reads /log/print on its own cadence regardless of whether you forward logs anywhere.

NetFlow

Configuration of the exporters on each device, and the collector built into the platform. See Traffic analytics for what is done with the data.

The Sessions switch on each device keeps one row per connection from its flows, on Security → Sessions. It is MikroTik Manager's own setting; nothing is written to the device.

Discovery and SNMP

Neighbour discovery protocol settings (LLDP, CDP, MNDP) per device, and SNMP configuration.

Discovery is what populates the topology map and finds devices to adopt. Turning it off on a device makes that device invisible to its neighbours — and this platform.

The LLDP section covers every online device in the current site: routers, switches, wireless APs and anything else running RouterOS. Use the tabs to narrow it to one type. Enable and Disable only touch the devices listed.

The API is GET /api/network-services/lldp and PUT /api/network-services/lldp with {"enabled": true, "device_types": ["wireless_ap"]}. Leave out device_types to target every type. The older /api/routers/lldp and /api/switches/lldp still work.

Applying SNMP settings

The SNMP section covers every online device in the current site: routers, switches, wireless APs and anything else running RouterOS (it used to cover routers and switches only). Use the tabs to narrow the list to one type.

Choose where a change goes with Choose devices (one or several, with Select all), then Apply to selected, or use Apply to all. Nothing is pushed automatically, and Apply asks you to confirm the exact changes and the devices they go to first.

The form shows what the chosen devices have now, or, with none chosen, what every listed device has. A setting they don't all share shows as differs (the slider sits in the middle) instead of a made-up default.

Only what you change is sent. The form remembers which fields you edited and highlights them; everything else is left as it is on each device, so setting a trap destination doesn't also rewrite each device's community or turn SNMP on or off. Reset discards your edits. A community name a device doesn't have is added next to its existing ones; nothing is renamed or removed. Before 0.24.37 the form was filled in from one device and every field was sent, so Apply copied that device's community, version and on/off state to the whole fleet, renaming each device's first community.

For SNMPv3, leaving the privacy password blank keeps the current one and keeps the user encrypted. It used to drop the user to authentication only.

The API is GET /api/network-services/snmp (every device type) and PUT /api/network-services/snmp, which needs device_ids: a list of device ids, or "all" (optionally with device_types). Only the settings in the body are changed, for example {"device_ids": [8, 12], "trap_target": "10.0.0.5"}. The older /api/routers/snmp and /api/switches/snmp still work.

SNMP contact and location per device

Contact, location and trap target can use variables, filled in per device:

Variable Value
{identity} RouterOS identity (/system identity). {$systemidentity} also works
{name} Device name in the manager
{ip} Management address
{model}, {serial} From the device
{site} Site name
{location} The device's location, or its site's address if it has none

For example, contact {identity}@example.com or location {site} / {location}. An unknown variable is rejected before anything is written.

The form remembers the last contact, location and trap destination you applied, variables and all, and shows them when you come back (the table shows each device's filled-in value). A remembered value isn't sent unless you change it or press Send it under the field, which is how you apply the same template to devices added later.

A blank field keeps each device's current value. Before 0.24.26 a blank field overwrote it with nothing.