Subscriber troubleshooting · free tool

PPPoE MTU and TCP MSS calculator

Work out where the bytes go before changing a subscriber’s MTU. Compare VLAN, QinQ, PPPoE and tunnel overhead, then copy RouterOS examples for review.

MikroTik integration · guided pilot

Build your encapsulation stack

All calculations stay in this browser. This estimates a byte budget; it does not measure your path.

Enter the Ethernet payload capacity, excluding the base 14-byte Ethernet header and 4-byte FCS. A normal 1500-byte IP link often has extra room for VLAN tags: do not subtract them twice. Tags here are on the physical access link, outside any tunnel.

Order: access tags → EoIP/VXLAN (if selected) → PPPoE (if selected) → L2TP/WireGuard (if selected) → subscriber IP. Only one tunnel is modelled.

Effective IP MTU
1492 bytes
PPPoE MTU / MRU ceiling
1492 / 1492
IPv4 TCP MSS
1452 bytes
IPv6 TCP MSS
1432 bytes

MTU and MRU are symmetric upper bounds for this model. Actual PPP negotiation, a peer limit, or a smaller hop can lower them. MSS uses fixed IP + TCP headers: 40 bytes for IPv4, 60 for IPv6.

Overhead per layer
LayerHeader bytesBudget used
PPPoE header66
PPP protocol field22

RouterOS review examples

Use the client or server example matching your setup. Profile MSS adjustment is IPv4 only. Routed tunnels need their own inner MTU and MSS handling. Mangle rules start disabled; choose one clamping approach, check both traffic directions and rule order, and test before enabling.

# Review examples; replace names and verify the negotiated MTU before applying.
# PPPoE client example. MTU/MRU are interface settings, not /ppp profile properties.
/interface pppoe-client set [find where name="REPLACE-PPPOE-CLIENT"] max-mtu=1492 max-mru=1492
# Server alternative: use the matching service only, instead of the client line.
/interface pppoe-server server set [find where service-name="REPLACE-SERVICE"] max-mtu=1492 max-mru=1492
# IPv4 profile option (alternative to the mangle rules below).
/ppp profile set [find where name="REPLACE-PROFILE"] change-tcp-mss=yes
# Set/verify the subscriber interface MTU at 1492 bytes.
# Mangle alternative: review rule order and FastTrack; enable only after testing.
# Select the interface facing the constrained link; cover SYNs in both directions.
# On a PPPoE server, each subscriber has a dynamic session interface; a single interface matches only one session.
# To cover those sessions, replace in-interface/out-interface with in-interface-list/out-interface-list and use a list containing the PPPoE session interfaces.
/ip firewall mangle add chain=forward out-interface="REPLACE-PPPOE-INTERFACE" protocol=tcp tcp-flags=syn action=change-mss new-mss=clamp-to-pmtu disabled=yes comment="MTU review"
/ip firewall mangle add chain=forward in-interface="REPLACE-PPPOE-INTERFACE" protocol=tcp tcp-flags=syn action=change-mss new-mss=clamp-to-pmtu disabled=yes comment="MTU review"
# PMTU clamping uses route/interface knowledge; it does not probe a hidden downstream bottleneck.

Subscriber-side packet test

Replace the documentation address with a reachable test host you control beyond the constrained link. These Linux commands use ICMP payload sizes (IP and ICMP headers excluded):

ping -4 -M do -c 3 -s 1464 192.0.2.1
# Then try one byte larger:
ping -4 -M do -c 3 -s 1465 192.0.2.1

IPv4 uses the DF (Don’t Fragment) bit. On Windows, use ping -f -l with the same payload size. Start smaller if this fails, then increase to find the largest working payload. A timeout alone does not prove an MTU problem: the host may filter ICMP. Test both directions and allow the relevant ICMP errors.

Why some sites load and others stall

Small DNS queries and TCP handshakes can fit through a link while larger packets fail. When an IPv4 packet has Don’t Fragment set, a smaller hop must return an ICMP fragmentation-needed error. If that error is filtered, the sender may keep retrying packets that cannot pass. IPv6 relies on ICMPv6 Packet Too Big messages and routers do not fragment packets. See RFC 2923 and RFC 8200.

MSS clamping can reduce the TCP data size negotiated in SYN packets. It does not repair the underlying path, change existing connections, or fix UDP/QUIC. Restore the required ICMP feedback, check the negotiated PPP limits, and retest the affected service. A clamp-to-pmtu rule uses the router’s route/interface information; it cannot discover an unknown smaller hop by itself.

Read the result at the right layer

EoIP and VXLAN carry Ethernet, so their headers come before the inner PPPoE session in this model. L2TP and WireGuard presets carry subscriber IP over the access link; when PPPoE is selected, its MTU is the tunnel’s underlay ceiling. The effective IP MTU is then smaller than the PPPoE MTU.

The model uses one tunnel, outer access tags, fixed IP headers, and no IPsec, nested tunnels, extra inner VLANs or compression. L2TPv2 is the minimum routed, unencrypted format. WireGuard reserves padding conservatively. If your topology differs, measure and account for those extra layers before applying any configuration.

RouterOS settings and validation

Set max-mtu and max-mru on the matching PPPoE client or server service. They are not PPP profile properties. The PPP profile’s change-tcp-mss setting applies to IPv4 only. For IPv6 or an inner routed tunnel, review the corresponding mangle rules and effective interface MTU. See PPPoE settings and PPP profiles.

The examples deliberately use replacement names and disabled mangle rules. Review the current configuration, choose one clamping method, test new connections, and preserve a rollback copy. Do not apply both client and server examples blindly. A successful ping to one destination does not establish the MTU to every destination.

Where ISPAgents fits

Automated ISPAgents MSS auditing is planned. This calculator does not inspect or change your router. MikroTik RouterOS operations are a guided pilot; guarded writes and RouterOS 6 fallback remain product work. Review the MikroTik integration or read the RADIUS deployment guide.

Frequently asked questions

Does a VLAN always reduce a 1500-byte IP MTU?

No. Many Ethernet interfaces have enough Layer 2 capacity for tags in addition to a 1500-byte payload. Subtract four bytes per tag only when the entered budget must include those tags. Check every hop.

What are the usual PPPoE MTU and TCP MSS values?

With a 1500-byte Ethernet payload and no other overhead, PPPoE uses eight bytes, leaving MTU 1492, IPv4 MSS 1452 and IPv6 MSS 1432. Negotiation or a smaller path can lower those values.

Can MSS clamping fix every broken site?

No. It changes TCP SYN negotiation only. UDP and QUIC still need correct MTU and path MTU discovery. Fix filtered ICMP errors and validate both directions.

Is this an ISPAgents router audit?

No. This is a local calculator with editable assumptions. Automated ISPAgents MSS auditing is planned; these results are not a router inspection or a deployment action.

Overhead sources

Byte values and assumptions: RouterOS MTU and VLANs, RFC 2516 (PPPoE), RFC 4638, EoIP, VXLAN, RFC 2661 (L2TP), WireGuard transport format, and RFC 6691 (MSS).

Browse ISP operations tools