test_runner: use LFC by default

[bayandin]

Problem

We don't currently test LFC enough, because we disable LFC by default.

Summary of changes

LFC was enabled by default, use the environment variable USE_LFC=false to disable.

Checklist before requesting a review

• I have performed a self-review of my code.
• If it is a core feature, I have added thorough tests.
• Do we need to implement analytics? if so did you add the relevant metrics to the dashboard?
• If this PR requires public announcement, mark it with /release-notes label and add several sentences in this section.

Checklist before merging

• Do not forget to reformat commit message to not include the above checklist

[github-actions]

6512 tests run: 6228 passed, 1 failed, 283 skipped (full report)

---

Failures on Postgres 17

• test_tenant_import[None-local_fs]: debug-x86-64

# Run all failed tests locally:
scripts/pytest -vv -n $(nproc) -k "test_tenant_import[debug-pg17-None-local_fs]"

Flaky tests (1)

Postgres 16

• test_timeline_archive[4]: release-arm64

Test coverage report is not available

_{The comment gets automatically updated with the latest test results
4eed077 at 2024-11-06T09:23:23.366Z :recycle:}

[koivunej]

I see no problem with enabling LFC by default, it's a great idea. Not extending the current amount of caching (1MB by default) might be the best way to not introduce flakyness.

Note we have some regress and perf tests which configure shared buffers. Perhaps those could also be using a larger LFC?

$ rg 'shared_buffers\s*=' test_runner/
test_runner/fixtures/compare_fixtures.py
196:                "shared_buffers=1MB",

test_runner/performance/test_compaction.py
38:        "main", tenant_id=tenant_id, config_lines=["shared_buffers=512MB"]
82:        "main", tenant_id=tenant_id, config_lines=["shared_buffers=512MB"]

test_runner/performance/test_startup.py
52:                    config_lines=["shared_buffers=262144"],

test_runner/performance/test_seqscans.py
55:            shared_buffers = row[0]

test_runner/performance/pageserver/pagebench/test_large_slru_basebackup.py
92:        "main", tenant_id=template_tenant, config_lines=["shared_buffers=1MB"]

test_runner/regress/test_lfc_working_set_approximation.py
21:            "shared_buffers='1MB'",
85:            "shared_buffers=1MB",

test_runner/regress/test_pageserver_catchup.py
15:        "test_pageserver_catchup_while_compute_down", config_lines=["shared_buffers=512MB"]

test_runner/regress/test_pg_query_cancellation.py
52:            "shared_buffers = 128MB",

test_runner/regress/test_local_file_cache.py
24:            "shared_buffers='1MB'",

test_runner/regress/test_wal_acceptor.py
2151:        config_lines=["shared_buffers=1MB"],

test_runner/regress/test_explain_with_lfc_stats.py
19:            "shared_buffers='1MB'"

[bayandin]

@a-masterov is going to continue the work on this PR

[bayandin]

Last Week:

• LFC made tests more flaky due to higher disk IO

This Week:

• Need to figure out what to do with disk
• Code review from @bayandin