[Store] Make batch_evict_bench scale and eviction ratios configurable

[jacklin78911-collab]

Description

Adds three optional gflags CLI flags to the BatchEvict benchmark so the 10M-objects / 5%-eviction workload discussed in the eviction-optimization work can be reproduced directly from the main-branch benchmark:

• --num_objects: run a single custom scale when greater than 0; 0 preserves the default scale list
• --evict_ratio_target: eviction target ratio (default 0.50)
• --evict_ratio_lowerbound: lower-bound ratio (default 0.25)

Switched from environment variables to gflags CLI flags per review.

Defaults are unchanged: without these flags, the benchmark runs the same scales (10k, 100k, +1M with MOONCAKE_EVICT_BENCH_LARGE=1) with the same ratios and identical CSV output. Invalid ratio combinations and invalid flag values fail with a non-zero exit. The existing lock-probe environment variables and their parsing are untouched. This is a benchmark-only change; no production code is modified.

Follow-up to the benchmark added in #2584; relates to the eviction-cycle discussion in #2560.

Module

• Mooncake Store (mooncake-store)

Type of Change

• Other (benchmark-only configurability change)

How Has This Been Tested?

Test commands:

# Default regression
./build/mooncake-store/benchmarks/batch_evict_bench

# Invalid ratio combination: exits non-zero
./build/mooncake-store/benchmarks/batch_evict_bench \
  --evict_ratio_target=0.05 \
  --evict_ratio_lowerbound=0.5

# Invalid non-numeric flag value: exits non-zero
./build/mooncake-store/benchmarks/batch_evict_bench \
  --num_objects=invalid

# Custom workloads
./build/mooncake-store/benchmarks/batch_evict_bench \
  --num_objects=1000000 \
  --evict_ratio_target=0.05 \
  --evict_ratio_lowerbound=0.05

./build/mooncake-store/benchmarks/batch_evict_bench \
  --num_objects=10000000 \
  --evict_ratio_target=0.05 \
  --evict_ratio_lowerbound=0.05

[gemini-code-assist]

Code Review

This pull request introduces dynamic configuration of eviction ratios and benchmark object counts via environment variables in the batch eviction benchmark. It adds helper functions to parse and validate these environment variables. The feedback highlights a critical issue in ReadStrictEnvSize where std::getenv can return nullptr, leading to undefined behavior when constructing a std::string_view without a prior null check.

Important

The consumer version of Gemini Code Assist on GitHub is being sunset. Starting June 18, 2026, new organization installations will be blocked, and all code review activity will officially cease on July 17, 2026.
For more details on the timeline and next steps, please review the Help Documentation.

[codecov-commenter]

⚠️ Please install the to ensure uploads and comments are reliably processed by Codecov.

Codecov Report

❌ Patch coverage is 0% with 18 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
mooncake-store/benchmarks/batch_evict_bench.cpp	0.00%	18 Missing ⚠️

📢 Thoughts on this report? Let us know!

[yokinoshitayoki]

Could we expose these three new benchmark parameters as gflags command-line options instead of environment variables, for example --num_objects, --evict_ratio_target, and --evict_ratio_lowerbound. This benchmark already uses gflags, and CLI flags are consistent with the other Store benchmarks, discoverable through --help, and avoid unnecessary custom environment-variable parsing.

[yokinoshitayoki]

Thanks for your contribution, this PR looks good to me now.