Skip to content

tests: fix flaky test_sql blockheight race#9304

Open
ksedgwic wants to merge 1 commit into
ElementsProject:masterfrom
ksedgwic:fix-flaky-test-sql-blockheight
Open

tests: fix flaky test_sql blockheight race#9304
ksedgwic wants to merge 1 commit into
ElementsProject:masterfrom
ksedgwic:fix-flaky-test-sql-blockheight

Conversation

@ksedgwic

Copy link
Copy Markdown
Collaborator

Observed on CI runs for #9299 and #9300 (Valgrind Test CLN 12/12):
https://github.com/ElementsProject/lightning/actions/runs/29053072126
https://github.com/ElementsProject/lightning/actions/runs/29053114743

The test mines ~106 blocks immediately before making three payments,
without waiting for the nodes to process them. Under valgrind the
paying node can lag far behind bitcoind, so it builds onions whose
cltv expiries the caught-up peers reject. The two runs showed both
shapes of the race:

  • l1 ~40 blocks behind (l2/l3 at 210): l2 failed the forward ("Expiry
    cltv 181 too close to current 210", expiry_too_soon). The error
    came from the invoice's route-hint channel, so xpay disabled it and
    the payment failed hard with error 209.

  • l1 a few blocks behind: l3 failed the final hop ("Expiry cltv too
    soon 211 < 210 + 5"). xpay's waitblockheight retry succeeded, but
    the failed first attempt left status 'failed' in the forwards row
    the test asserts is 'settled'.

The test paid with pay until 7f6a333 switched it to xpay; pay took
its start height from bitcoind's headercount and tracked the node's
chainlag, so lagging nodes still built onions valid at the true chain
tip, which is why this only started flaking recently.

The fix syncs the nodes' blockheights before paying, as other tests
do after mining bursts.

Changelog-None

The test mines ~106 blocks (channel close, replacement channel
funding, then generate_block(99)) immediately before making three
payments, without waiting for the nodes to process those blocks.
Under valgrind the paying node can lag far behind bitcoind when the
payments start, so it builds onions whose cltv expiries the caught-up
peers reject.  Two CI failures showed both shapes of the race:

In one run l1 was ~40 blocks behind (l2/l3 at block 210, l1 still
catching up), so l2 failed the forward: "Expiry cltv 181 too close to
current 210" (expiry_too_soon).  Since the error came from the
invoice's route-hint channel, xpay disabled it and the payment failed
hard: "All 1 channels to the destination are disabled."

In the other run l1 was only a few blocks behind, so l3 failed the
final hop: "Expiry cltv too soon 211 < 210 + 5".  xpay's
waitblockheight retry then succeeded, but the failed first attempt
left status 'failed' in the forwards row the test later asserts is
'settled'.

The test paid with pay until 7f6a333 ("pytest: use xpay, not pay in
misc tests."); pay took its start height from bitcoind's headercount
and tracked the node's chainlag, so a lagging node still built onions
valid at the true chain tip.  xpay uses its own notification-fed
blockheight plus one block of slack, which is why this only started
flaking recently.

Sync the nodes' blockheights before paying, as other tests do after
mining bursts.

Changelog-None
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant