Merged
Conversation
# Description This is a some kind of follow-up to cowprotocol#3578, where I noticed that liquidity tick net is not used when working with concentrated univ3 liquidity sources in the Baseline solver, while this data might be heavy and contain thousands of records for each pool, which can potentially lead to unreasonable memory consumption. # Changes Do not store the liquidity tick net received from driver. It would be great to not even parse it, but a shared struct is being used here, so the parsed auction is now consumed by the `into_domain` function to be dropped as soon as possible to avoid storing it in memory for the whole `solve` operation. ## How to test No tests so far for univ3 liquidity due to annoying issues with its fetching delays in e2e tests.
# Description Even after cowprotocol#3499, a memory leak [was noticed](cowprotocol#3554) when the UniV3 liquidity fetching is enabled in the Baseline solver. Looking at the code, I don't see an obvious reason for it other than the UniswapV3QuoterV2 contract instance is being extensively cloned on each `/solve` request. # Changes - Use `Arc`'ed UniswapV3QuoterV2 contract instance. ## How to test I've tried to resurrect [this](cowprotocol@a9ff88f) e2e test, but the liquidity fetching form subgraph sometimes takes a very long time, so the test is too flaky to enable it. ## Follow-ups In any case, I need to find a way to make the Jemalloc profiler work[cowprotocol#3533] to collect memory dumps later in case the memory leak happens again. ## Related Issues cowprotocol#3554
# Description cowprotocol#3532 set persist-credentials to false leading the to the tag not being pushed as there is no github creds stored to do so.
This PR adds Trivy (A Docker vulnerability scanner) to the repository. --------- Co-authored-by: ilya <ilya@cow.fi> Co-authored-by: Martin Magnus <martin.beckmann@protonmail.com>
# Description We fixed foundry version for driver tests in cowprotocol#3561 because high_gas_limit started failing. This PR fixes the test with the latest foundy version and adds comments that make it clear what we're testing. I also added a test that makes sure we discard solutions that take more gas than half of the block gas limit.
# Description For an order to be put into the current auction the autopilot needs to know the native price of the sell and the buy token. Any delays in fetching the native prices can result in orders having "to sit out" an auction which immediately adds at least 12s to the execution time. But we also have to be mindful of the time we spend fetching native prices because that delays the entire auction. Also in order to reduce cost we fetch native prices in stages (first the cheap quoters, then the expensive ones). # Changes So far we evenly divided the time for each stage due to its simplicity. However, we are currently dealing with issues where the later stages just need a little bit more time. To get the most out of the allocated time and use it fully this PR adjusts the way we compute the timeout for the current stage. Now the timeout of the current stage will adjust dynamically. The current stage always gets `total_time_left / stages_left`. This does not change anything for the very first stage but subsequent stages would get more time whenever some stage does not use the entire allocated time budget. For the native price estimation code to know how many stages are left we build a new `Context` struct in the `CompetitionEstimator` that contains that information. I also `async-ified` the other tests which turned out to be unnecessary. I think the new way is cleaner but LMK if you prefer to have that in a separate PR. ## How to test Added a unit test to assert that the timeouts are what we'd expect them to be (left code comments to explain the subtle details)
# Description Follow up PR from cowprotocol#3569. > [!CAUTION] > Given the parsing logic is changed this PR will need to be synchronized with an infrastructure change upon merge. # Changes - [x] reduces the number of choices we have to configure our gas estimators in the API and autopilot to reflect current reality (GasNot, Blocknative, etc don't really exist anymore and are not used in prodcution) - [x] changes the configuration to not require two orthogonal parameters (gas_estimator & gas_estimator_driver_url) and clumsy logic to ensure consistency and collapses them into a single type that can be parsed - [x] remove some otherwise unneeded code ## How to test Existing tests
) # Description In cowprotocol#3542 we improved the performance of loading live orders on `autopilot` startup, but it turns out that this made the performance of counting orders per user worse: [grafana](https://g-0263500beb.grafana-workspace.eu-central-1.amazonaws.com/explore?schemaVersion=1&panes=%7B%22zj9%22:%7B%22datasource%22:%225An5Ejq4k%22,%22queries%22:%5B%7B%22refId%22:%22A%22,%22expr%22:%22rate%28gp_v2_api_orderbook_database_queries_sum%7Btype%3D%5C%22count_limit_orders_by_owner%5C%22,%20network%3D%5C%22mainnet%5C%22%7D%5B5m%5D%29%20%2F%20rate%28gp_v2_api_orderbook_database_queries_count%7Btype%3D%5C%22count_limit_orders_by_owner%5C%22,%20network%3D%5C%22mainnet%5C%22%7D%5B5m%5D%29%22,%22range%22:true,%22instant%22:true,%22datasource%22:%7B%22type%22:%22prometheus%22,%22uid%22:%225An5Ejq4k%22%7D,%22editorMode%22:%22builder%22,%22legendFormat%22:%22__auto%22,%22useBackend%22:false,%22disableTextWrap%22:false,%22fullMetaSearch%22:false,%22includeNullMetadata%22:true%7D%5D,%22range%22:%7B%22from%22:%22now-6h%22,%22to%22:%22now%22%7D%7D%7D&orgId=1) This PR bring the old query back for limit order counting which se use [in the orderbook](https://github.com/cowprotocol/services/blob/main/crates/shared/src/order_validation.rs#L712)
# Description
Some of our logs have become pretty hard to read IMO.
Take this log for example:
`
2025-08-27T07:38:26.694Z INFO autopilot_startup{args.shared.node_url=Url
{ scheme: "http", cannot_be_a_base: false, username: "", password: None,
host: Some(Domain("mainnet-proxy.rpc-nodes.svc.cluster.local")), port:
None, path: "/", query: None, fragment: None
}}:ethereum{chain=Mainnet}:current_block_stream: ethrpc::block_stream:
noticed a new block number=23231015
hash=0xa65cfa24ed5be0f2f3d8c63639f423c4dcaaa50ebeba05f218349d5be8f30ebd
trace_id=83ee7db2a5fdfac5e413ffbb4eba4652
`
AFAICS roughly half of the log is not really useful
(`args.shared.node_url=Url { scheme: "http", cannot_be_a_base: false,
username: "", password: None, host:
Some(Domain("mainnet-proxy.rpc-nodes.svc.cluster.local")), port: None,
path: "/", query: None, fragment: None }}:ethereum{chain=Mainnet`).
# Changes
1. stop logging the node_url on every log during autopilot startup
2. stop logging the current network as this is already part of the
elasticsearch document
3. added startup tracing for driver and orderbook as I think it's
overall nice to easily see all logs that happen during the startup
sequence
@fafk will this interfere with the functionality of `tempo`? I'm mostly
concerned about `2` in case it's somehow used to aggregate data at the
moment.
# Description Pancake v2 pool digest on BNB is different. I found the correct one by going to https://bscscan.com/address/0xca143ce32fe78f1f7019d7d551a6402fc5350c73#readContract and reading the `INIT_CODE_PAIR_HASH` constant. I was able to get a quote with this hash.
|
I have read the CLA Document and I hereby sign the CLA 0 out of 7 committers have signed the CLA. |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
7 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Description
Changes
How to test