construction: switch to SubmitTxNoWait

[pro-wh]

fixes #24

switch to this new API in 20.9

[pro-wh]

what's with this submission?

(gateway) :55:14.443 - ConstructionHash OK for 9f7f67e02e80dec4b4c35d84d61adf838dc5a727ec5a28444865fadd80125efc

(cli) Transaction Created: 9f7f67e02e80dec4b4c35d84d61adf838dc5a727ec5a28444865fadd80125efc
oasis1qpxnnlakg83xu8c2xw9jrz05mh0733twwq54xjga -- 54.387348712 ROSE --> oasis1qzm8zn43fl9lc20pr4gkkdks5htwu3r8wucz38dv

(gateway) :55:14.446 - ConstructionSubmit OK for 9f7f67e02e80dec4b4c35d84d61adf838dc5a727ec5a28444865fadd80125efc

(cli) :55:19 - {Code:15 Message:unable to submit transaction Retriable:false Details:map[]}: unable to broadcast transaction: unable to broadcast transaction 9f7f67e02e80dec4b4c35d84d61adf838dc5a727ec5a28444865fadd80125efc

(cli) :55:30 - {Code:15 Message:unable to submit transaction Retriable:false Details:map[]}: unable to broadcast transaction: unable to broadcast transaction 9f7f67e02e80dec4b4c35d84d61adf838dc5a727ec5a28444865fadd80125efc

(gateway) :55:40.154 - Block OK containing 9f7f67e02e80dec4b4c35d84d61adf838dc5a727ec5a28444865fadd80125efc

(cli) Check failed: broadcast failed for transaction 9f7f67e02e80dec4b4c35d84d61adf838dc5a727ec5a28444865fadd80125efc

why is there a ~5 second delay after ConstructionSubmit OK and then the CLI says it's unable to submit transaction?

[pro-wh]

maybe they rebroadcast it after a while even if the first broadcast succeeded

[pro-wh]

alright, they for sure rebroadcast it if the previous attempt succeeded. even after #40, it still fails due to too many broadcast attempts. all we have to do is raise the limit or make the interval slower

[pro-wh]

wow how far would we have to increase it? in https://github.com/oasisprotocol/oasis-core-rosetta-gateway/pull/38/checks?check_run_id=986970053, the transaction with hash 4f97e85bd6f7cdcf8d0e837fe844de45cb1416acd69027a10c021b63b09ff9b0 is created at ~:41:30 and submitted. but it doesn't appear in a block until :42:59!

[pro-wh]

hm, during that ~minute and a half, the block handling goes from height 1209 to 1501. that seems like a lot. is the block production just running away, leaving the gateway in the dust?

[kostko]

Ohh, it seems that it uses an old fixture format (see tests/test-fixture-config.json) so the commit timeout is not configured properly and it probably uses the insanely fast default. And this is something that changed in 20.9 indeed.

Instead of:
https://github.com/oasisprotocol/oasis-core-rosetta-gateway/blob/610438f34a10c11fe77048ded8464b206bf0b90d/tests/test-fixture-config.json#L6-L9

It should be:

    "network": {
        "node_binary": "oasis-node",
        "consensus": {
            "backend": "",
            "params": {
                "timeout_commit": 1000000000,