should construction be less flaky?

[pro-wh]

if you run rosetta-cli check:construction indefinitely, you get a failure like this:

[STATS] Transactions Confirmed: 127 (Created: 137, In Progress: 10, Stale: 0, Failed: 0) Addresses Created: 72
Check failed: Transaction Conflict. Please retry: unable to add block to storage e75e4f272b4cd8df8a2b897e011e4e67fdeac22177c63b8bbf7fdc9f70850ae1:302: unable to sync to 302

how bad is that?

cross reference #26 (comment)

[kostko]

Huh, what does the error mean?

[pro-wh]

they do this thing where in some places the wrap errors with the style "(cause): (context)" whereas we (outside of this repo) often do it the other way around. but they also sometimes to "(context): (cause)," as seen in that "Check failed: ..."

[kostko]

Sure, but why is there a "transaction conflict" and why is there a "please retry" (is it a transient error?) and why is it "unable to add block to storage" and "unable to sync"?

[pro-wh]

that "Transaction Conflict. Please retry" seems to come from badger https://github.com/dgraph-io/badger/blob/v2.0.3/errors.go#L43

[kostko]

This looks like an upstream bug, they should correctly handle badger tx conflicts.