Libreswan's Online Documentation
View the Project on GitHub libreswan/libreswan.github.io
SupportHere’s several suggestions on how to handle retransmits and timeouts. The thing to remember is that the transport essentially 100% reliable so retransmits are not needed.
... # sanitize-retransmits to commands ¶
Removes retransmits from just that commands sanitized output, so expect to apply it liberally.
conn %default; retransmit-timeout=10s; retransmit-interval=20s ¶
Since most things should finish in under 10s there should never be a retransmit.
--impair timeout:5 is proposed as an alternative
conn %default; retransmit-timeout=5s; retransmit-interval=10s ¶
For when a timeout is expected. Presumably other exchanges complete in less than 5s.
--impair timeout:5 is proposed as an alternative
conn %default; retransmit-timeout=9s; retransmit-interval=5s ¶
IKEv1 sends the full payload, and then retransmits with a fragmented payload. The above makes that visible.
--impair timeout_on_retransmit ¶
It only worked reliably when the first message was expected to timeout.
When multiple exchanges are involved (for instance IKE_INTERMEDIATE), it could easily timeout for no reason other than the peer taking longer than usual.
The problem was that it sanitized retransmits when they were part of the test.