-
Notifications
You must be signed in to change notification settings - Fork 77
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
fix(composer)!: update to work with appside mempool #1643
Conversation
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Please also add a black box test for this new error. See existing test for INVALID_NONCE
:
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
The abci error codes are only one part in addressing #1633.
This patch also needs to change how composer fetches the next nonce to use (since right it makes use of cometbft's get_nonce
, which is not aware of the mempool).
note: not ready for review yet |
c0b80ae
to
f6c5a05
Compare
@@ -8,7 +8,8 @@ data: | |||
ASTRIA_COMPOSER_API_LISTEN_ADDR: "0.0.0.0:{{ .Values.ports.healthApi }}" | |||
ASTRIA_COMPOSER_GRPC_ADDR: "0.0.0.0:{{ .Values.ports.grpc }}" | |||
ASTRIA_COMPOSER_SEQUENCER_CHAIN_ID: "{{ tpl .Values.config.sequencerChainId . }}" | |||
ASTRIA_COMPOSER_SEQUENCER_URL: "{{ tpl .Values.config.sequencerRpc . }}" | |||
ASTRIA_COMPOSER_SEQUENCER_HTTP_URL: "{{ tpl .Values.config.sequencerRpc . }}" |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
we call this COMETBFT_ENDPOINT
in other places
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
updated the name in the code as well
}) | ||
.with_config(retry_config) | ||
.await | ||
.wrap_err("failed getting latest nonce from sequencer after 1024 attempts"); | ||
.wrap_err("failed getting pending nonce from sequencing after 1024 attempts"); |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
"from sequencer" instead of "from sequencing"?
@@ -127,18 +99,57 @@ async fn invalid_nonce_causes_resubmission_under_different_nonce() { | |||
|
|||
// wait for 1 sequencer block time to make sure the bundle is preempted | |||
tokio::time::timeout( | |||
Duration::from_millis(test_composer.cfg.block_time_ms), | |||
Duration::from_millis(test_composer.cfg.block_time_ms + 1000), |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
should probably remove the + 1000
here?
} | ||
} | ||
|
||
pub(crate) async fn mount_pending_nonce_response( |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
this should probably return a mock guard
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
It's fine if it doesn't. Ideally you provide the intermediate fetches (maybe with an expect4ed number of requests that match) and only have a mockguard on the actual output of the service (in this case, a transaction that contains a collection of sequence actions).
If you return a mock guard and then wait_until_satisfied
it, you end up in a situation where you test the implementation (in which order are requests made?) rather than its behavior (does it produce the expected result?).
let expected_rollup_ids = vec![RollupId::from_unhashed_bytes("test1")]; | ||
// Expect nonce 1 again so that the resubmitted tx is accepted | ||
let valid_nonce_guard = | ||
mount_broadcast_tx_sync_mock(&test_composer.sequencer, expected_rollup_ids, vec![1]).await; | ||
|
||
// Mount a response of 1 to a nonce query | ||
test_composer |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
should get a mock guard from the mount helper and then wait for it to be satisfied
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Noted elsewhere: ideally we don't worry about the choreography of which RPCs are made in which order. It's best to just provide them and then try and observe what comes out of the service. Makes the tests less fragile.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
the pending nonce mock should use a mockguard and wait_until_satisfied
as the mount is used several times in the test (e.g. for initializing nonce on startup and then on refetch after the invalid/taken nonce errors). besides that, just a couple nits on the error message and config variable names.
bd37417
to
f0e7f34
Compare
f0e7f34
to
cfbf479
Compare
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Went over the deleted tests and made an issue for adding the coverage back as black box tests: #1652
/// Test to check that the executor sends a signed transaction to the sequencer as soon as it | ||
/// receives a `SequenceAction` that fills it beyond its `max_bundle_size`. | ||
#[tokio::test] | ||
async fn full_bundle() { |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This test does not have an analogous black box test currently. However, there are unit tests on the BundleFactory
that test for full bundles being fired off.
/// Test to check that the executor sends a signed transaction to the sequencer after its | ||
/// `block_timer` has ticked | ||
#[tokio::test] | ||
async fn bundle_triggered_by_block_timer() { |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This test also doesn't have an analog in the black box tests. We've tested this internally to the crate, in the executor module instead as this test assumes quite a bit of knowledge about the internal functions of the composer (specifically that if the bundle isn't filled then it will be submitted after the block timer).
/// Test to check that executor's chain ID check is properly checked against the sequencer's chain | ||
/// ID | ||
#[tokio::test] | ||
async fn chain_id_mismatch_returns_error() { |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This test does not currently exist in the blackbox tests as it requires bubbling errors up out from the composer service.
/// Test to check that the executor sends a signed transaction with two sequence actions to the | ||
/// sequencer. | ||
#[tokio::test] | ||
async fn two_seq_actions_single_bundle() { |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This test also does not have an analog blackbox test currently. A test similar to the existing single_rollup_tx_payload_integrity
that has two transactions should be added to cover this.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Thanks for the fix!
The tests in the executor
module are in a weird state and should be part of the blackbox tests instead. As they are testing relatively low impact functionality as per @itamarreif's notes, I am in favor of removing them here and following up in a future patch.
@@ -30,6 +29,8 @@ data: | |||
OTEL_EXPORTER_OTLP_TRACE_HEADERS: "{{ tpl .Values.otel.traceHeaders . }}" | |||
OTEL_SERVICE_NAME: "{{ tpl .Values.otel.serviceName . }}" | |||
{{- if not .Values.global.dev }} | |||
ASTRIA_COMPOSER_SEQUENCER_ABCI_ENDPOINT: "{{ tpl .Values.config.sequencerRpc . }}" |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
when .Values.global.dev
sets to true it marks a breaking change to configs. Therefore, new env variables should be added to else
bracket while values being removed should be added to {{- if not .Values.global.dev }}
bracket.
baac1c3
to
5bdfddc
Compare
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Charts LGTM, small nit
Co-authored-by: quasystaty <[email protected]>
Summary
Updates composer to work with the new appside mempool. New ABCI error codes were added and now nonces can be stacked, so the submission logic should use the GRPC
pending_nonce()
instead of the RPCget_latest_nonce()
.Background
New ABCI error codes were added in PR #1515 which need to be handled in the composer's submission logic.
New Error codes:
NONCE_TAKEN
-> try resubmitting with new top-of-pending. If this is happening because the composer is out of funds, that is okay as the composer has it's own fund monitoring logicPARKED_FULL
-> normal error case okACCOUNT_SIZE_LIMIT
-> normal error case okALREADY_PRESENT
-> normal error case ok, shouldn't be returnedBackground
We changed the internals of the mempool which resulted in the users of the mempool to have changes as well.
Testing
Smoke tests
Related Issues
closes #1633