Failures and antipatterns · Intermediate

MTU black holes

Test a path-MTU hypothesis by relating transfer size, required feedback and sender behavior rather than relying on one failed probe.

Ask Logeia about this articleTry the three questionsReview your answers

Updated

On this page
  1. Small exchanges work while larger transfers stall
  2. Path MTU and a missing feedback signal
  3. Vary packet size to test the hypothesis
  4. Handle required feedback and discovery
  5. Ask Logeia about this articleCheck your understanding

Small exchanges work while larger transfers stall

An MTU black hole occurs when packets exceed a path constraint and the sender does not receive or correctly use the feedback needed to reduce their size. A connection may establish and small requests may work while larger transfers repeatedly retransmit and stall. That pattern is a clue, not a complete diagnosis: other failures can also affect large responses. Test whether the boundary between success and failure tracks packet size. Application body size and individual IP packet size are different quantities, especially once segmentation and encapsulation enter the path.

Path MTU and a missing feedback signal

Suppose a sender uses 1500-byte IP packets but a route containing a tunnel can carry only 1400 bytes. With IPv4, a packet marked not to fragment may be dropped; if the needed ICMP feedback fails to reach the sender, retransmissions can repeat the same mistake. IPv6 routers do not fragment packets in transit, making Packet Too Big feedback important there as well. These sizes are an illustrative assumption rather than a measurement of a particular tunnel. Account for headers and encapsulation when determining the relevant path constraint.

Vary packet size to test the hypothesis

Compare small and large exchanges, and vary size or fragmentation-related options only within an authorized diagnostic environment. Packet captures, tracepath and sender retransmission information can help establish the sequence. A diagnostic ICMP probe need not receive the same treatment as the application’s TCP flow, so one failed large ping does not settle the cause. Examine VPNs, tunnels, load balancers and asymmetric paths. The useful evidence connects packet size, the drop point or missing feedback, and the sender’s subsequent behavior rather than merely observing that a download once failed.

Handle required feedback and discovery

Remedies can include permitting required ICMP feedback, configuring consistent tunnel MTUs and using suitable packetization-layer MTU discovery mechanisms. TCP MSS adjustment can mitigate some TCP paths but does not repair every UDP or encapsulation problem. Avoid globally suppressing ICMP or applying an unnecessarily tiny MTU to all hosts without considering security and performance consequences. Match the response to the affected layer and implementation, then repeat the same representative transfer. A change that lets one small test succeed is not enough to establish that larger packets or different directions now behave correctly.

Reading and practice are recorded separately in this browser, for each language.

Learn this in order

Understand the systems behind the web · STEP 10

Choose a learning path