wb | workbench improvements

[fmaste]

Description

This PR enhances the workbench profiling capabilities, streamlines Haskell dependency management, and introduces support for LSM-tree benchmarking profiles.

Key Changes

Multi-Dimensional Profiling Configuration

• Decoupling Build Type from Profiling Logic: Refactored the workbench to treat "Profiled Builds" and "Profiling RTS Parameters" as independent dimensions. Previously, RTS parameters were only accepted when a profiled build was active.
• WB_PROFILING Syntax: The WB_PROFILING environment variable now supports a new syntax: MAIN_PARAM+MODIFIER.
• • Main Parameters: Supports standard GHC RTS flags like -hc, -hi, -hT, or -p.
• • Modifiers: Supports eventlog and info-table.
• Increased Flexibility: This allows for granular control (e.g., enabling the eventlog on a non-profiled build and most important with info-table support) and combinations that were previously restricted. This significantly expands the debugging and analysis capabilities of the workbench.
• Examples: WB_PROFILING="space-heap" make ci-test, WB_PROFILING="space-heap" make ci-test-prof, WB_PROFILING="space-info+eventlog+info-table" make ci-test, WB_PROFILING="space-heap+eventlog" make value-volt-lsmt-nomadperf.

Enhanced Profiling Support

• Extended Shutdown Grace Period: Increased the wait time before stopping nodes in local clusters when profiling is enabled. This ensures the node has sufficient time to flush and dump all profiling data to disk before termination.
• Workbench Shell Integration: Profiling tools are now automatically included in the workbench shell environment.
• Selective Executable Profiling: In useCabalRun mode, profiling support is now explicitly scoped to cardano-node, cardano-tracer, tx-generator and locli. For *-nix profiles, the configuration will default to the settings defined in the haskell.nix project.
• RTS Options: Added missing Haskell Runtime System (RTS) profiling options and their descriptions (sourced from GHC documentation) as comments for better maintainability and discoverability.

Workbench Cleanup & Nix Integration

• Refactored Startup Logic: Removed redundant functionality and dead code from lib-cabal.sh. The workbench now relies on the Nix-generated start.sh script to pass RTS parameters, ensuring a more consistent environment across different execution modes (see how the different variables used to behave with the different combinations).
• Standardized Dependency Resolution: Refactored how the workbench obtains Haskell dependencies. By moving from flake-level outputs to a dedicated haskell.nix attribute set we achieve two major benefits:
• • Cache Stability: No need for careful use of .passthru.noGitRev to avoid "cache miss on every commit" issue caused by executables being patched with the GitHub commit hash. Now just haskellProject.exes.tx-generator.
• • Flexibility: It provides an independent haskell.nix attrset, allowing for future improvements without polluting the main flake logic.
• Simplified Parameter Passing: Refactored internal workbench Nix functions to stop passing both pkgs and lib as separate arguments. Since lib is a standard attribute of the pkgs set, functions now consume pkgs and access pkgs.lib internally. This reduces boilerplate and improves the consistency of the workbench Nix API.
• Removal of Legacy Keep-Alive Loop: Removed the keep-alive loop from Hydra's workbench-ci-test. This logic, which dates back to early 2022 (PR, commit), was originally implemented to prevent timeouts during long-running tests. After tracing its history and evaluating current CI behavior, it has been determined that this workaround is no longer necessary and was removed to simplify the test entry point.

New Benchmarking Profiles (LSM-Tree)

• LedgerDB.V2LSM Configuration: Introduced utxo_lsmt attributes to cardano-profile. This provides the necessary configuration parameters for LedgerDB.V2LSM, enabling the node to correctly initialize and test the LSM-based ledger database.
• LSM Tree Profile Support: Added new workbench profiles specifically for LSM trees. These are modeled after existing LMDB profiles to allow for direct performance comparisons between different backing stores.

How make behaved with and without profiling:

Makefile command	./shell.nix project	./nix/workbench/shell.nix	./lib-cabal.sh	start-cluster	start.sh
make ci-test	cardanoNodeProject	WB_PROFILING=	WB_PROFILING= WB_RTSARGS=	cabal_mode=t WB_PROFILING= WB_RTSARGS=	-
make ci-test-prof	ERROR: "WB_PROFILING must be one of: none, time, ..."	-	-	-	-
WB_PROFILING=space-heap make ci-test	cardanoNodeProject	WB_PROFILING=space-heap	WB_PROFILING=space-heap WB_RTSARGS=	cabal_mode=t WB_PROFILING=space-heap WB_RTSARGS=	-
WB_PROFILING=space-heap make ci-test-prof	cardanoNodeProject.profiled	WB_PROFILING=space-heap	WB_PROFILING=space-heap WB_RTSARGS=	cabal_mode=t WB_PROFILING=space-heap WB_RTSARGS=	-hT -machine-readable -tcardano-node.stats -pocardano-node
make ci-test-nix	cardanoNodeProject	WB_PROFILING=	-	cabal_mode=	-
make ci-test-profnix	ERROR: "WB_PROFILING must be one of: none, time, ..."	-	-	-	-
WB_PROFILING=space-heap make ci-test-nix	cardanoNodeProject	WB_PROFILING=space-heap	-	cabal_mode=	-
WB_PROFILING=space-heap make ci-test-profnix	cardanoNodeProject.profiled	WB_PROFILING=space-heap	-	cabal_mode=	-hT -machine-readable -tcardano-node.stats -pocardano-node

Checklist

• Commit sequence broadly makes sense and commits have useful messages
• New tests are added if needed and existing tests are updated. These may include:
  • golden tests
  • property tests
  • roundtrip tests
  • integration tests
    See Runnings tests for more details
• Any changes are noted in the CHANGELOG.md for affected package
• The version bounds in .cabal files are updated
• CI passes. See note on CI. The following CI checks are required:
  • Code is linted with hlint. See .github/workflows/check-hlint.yml to get the hlint version
  • Code is formatted with stylish-haskell. See .github/workflows/stylish-haskell.yml to get the stylish-haskell version
  • Code builds on Linux, MacOS and Windows for ghc-9.6 and ghc-9.12
• Self-reviewed the diff

Note on CI

If your PR is from a fork, the necessary CI jobs won't trigger automatically for security reasons.
You will need to get someone with write privileges. Please contact IOG node developers to do this
for you.

[mgmeier]

LGTM - impressive and very thorough. Thank you @fmaste