Expand SLO workload suite#64
Merged
Merged
Conversation
This was referenced Jun 16, 2026
Move the native query workload under slo-workload, add a shared core harness, and add JDBC and Spring-based workloads so all of them emit one consistent SLO metrics contract. Rebased onto master after #58 to drop the duplicated JDBC workload commits. Co-authored-by: Cursor <cursoragent@cursor.com>
KirillKurdyukov
force-pushed
the
slo-workload-suite
branch
from
c58781a to
cfac43b
Compare
Critical: - Add --shutdown-time, --max-attempts, --max-workers CLI flags; alias --duration as --time so ydb-slo-action invocations don't drop the flag - Reject unknown CLI options so typos in the action invocation fail loudly instead of silently falling back to defaults - Take an atomic snapshot of latency histograms (copy + reset) so concurrent recordValue can't tear p50/p95/p99 percentile readings - Align query workload's hash to the same client-side SplitMix64 used by the JDBC and Spring implementations so tables are byte-compatible across all workloads - Plumb retry counts through OpOutcome for the Spring workloads via a thread-local counter incremented inside @YdbTransactional methods, so sdk_retry_attempts_total isn't perpetually flat for those implementations - Size Hikari pool to 130 (2 × default --max-workers) so Spring workloads measure the JDBC driver, not pool contention; override with SLO_HIKARI_POOL_SIZE when raising --max-workers - Use Hibernate FlushMode.COMMIT and READ_ONLY hint on JPA reads, and EntityManager.clear() after each upsert, so the persistence context stops inflating read latencies and leaking memory Important: - Abort setup() if more than half the prefill rows fail or if every prefill worker fails to open a session, so reads never silently target an empty key-space - Decide retryable errors via YdbStatusable.getStatus().getCode().isRetryable on idempotent operations; fall back to the JDBC marker types only when the status isn't available - Treat SQLNonTransientConnectionException as a connection-level error so dead connections are reopened before the next attempt - Bound rate limiter acquire by the run deadline (tryAcquire with the remaining nanos) so a chaos stall during shutdown doesn't issue one extra operation past the deadline - Break the JDBC retry loop on interrupt instead of looping back into a failing executeQuery, so graceful shutdown doesn't inflate sdk_errors_total - Walk the cause chain in Spring classifyError looking for SQLState so Hikari pool exhaustion etc. surface as distinct error_kind labels instead of collapsing into "cannotgetjdbcconnectionexception" - DROP TABLE IF EXISTS in all workloads so a failed createTable doesn't produce a misleading teardown warning - Refactor Spring KvClient.forTable to return a fresh KvClient bound to the table path instead of mutating a singleton field - Add Span.finishAbortedIfOpen() safety net so pendingOperations never leaks if an unexpected throwable escapes the operation closure; volatile + AtomicIntegerFieldUpdater so double-finish can't double-decrement - Sanitize leading digits and cap length in Launcher.tablePathFor so YDB table-name constraints never reject the workload path Hygiene: - Extract DDL/DML templates into core.kv.KvSchema so the four implementations can't drift - Remove the unused JdbcTemplate bean from the JPA configuration - Update READMEs to match the new layout and document the new flags and the cross-implementation hash compatibility Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Gate child modules and release=17 behind a jdk17 profile so Java 8/11 jobs no longer hit --release; trim review leftovers and comments. Co-authored-by: Cursor <cursoragent@cursor.com>
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 join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
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.
Summary
slo-workload/query.slo-workload/coreharness so all workloads emit the same OTLP SLO metric contract.Test plan
mvn -B -q -pl slo-workload/core,slo-workload/query,slo-workload/jdbc,slo-workload/spring-data-jdbc,slo-workload/spring-data-jpa -am -DskipTests packageMerge order
Merge this PR first. The ydb-java-sdk and ydb-jdbc-driver workflow PRs depend on these paths existing on
master.Made with Cursor