fix(compute): Respect skip_pg_catalog_updates in reconfigure()

[ololobus]

Problem

We respect skip_pg_catalog_updates at the initial start, but ignore at the follow-up /configure. Yet, it's used for storage->cplane->compute notify requests after migrations, shard split, etc. So every time we get them, applying the new config takes much longer than it should because we go through Postgres catalog checks. Cplane sets this flag, when it does serves notify attach call https://github.com/neondatabase/cloud/commit/9068c7d7433f943af2bc350e9fd59772867e622c

Related to inc-403, for example

Summary of changes

Look at skip_pg_catalog_updates in compute.reconfigure()

[hlinnaka]

Makes sense. Does the control plane set skip_pg_catalog_updates correctly when reconfiguring for pageserver changes? And does it correctly not set it, when applying other changes?

[ololobus]

Makes sense. Does the control plane set skip_pg_catalog_updates correctly when reconfiguring for pageserver changes?

Yeah, see this commit https://github.com/neondatabase/cloud/commit/9068c7d7433f943af2bc350e9fd59772867e622c

And does it correctly not set it, when applying other changes?

From what I see, the logic is 'apply unless the flag is set' and it's not set by default, so I'd say yes

[github-actions]

7425 tests run: 7041 passed, 32 failed, 352 skipped (full report)

---

Failures on Postgres 17

• test_pageserver_compaction_circuit_breaker: release-x86-64-without-lfc, release-x86-64-with-lfc, release-arm64-with-lfc, release-arm64-without-lfc
• test_sharding_backpressure: release-x86-64-with-lfc, release-x86-64-without-lfc, debug-x86-64-without-lfc, release-arm64-with-lfc, release-arm64-without-lfc
• test_sharding_gc: release-x86-64-with-lfc, release-x86-64-without-lfc, release-arm64-with-lfc, release-arm64-without-lfc
• test_sharding_smoke: debug-x86-64-without-lfc, release-arm64-with-lfc, release-arm64-without-lfc

Failures on Postgres 16

• test_pageserver_compaction_circuit_breaker: release-x86-64-with-lfc, release-arm64-with-lfc
• test_sharding_smoke: release-x86-64-with-lfc, release-arm64-with-lfc
• test_sharding_backpressure: release-x86-64-with-lfc, release-arm64-with-lfc

Failures on Postgres 15

• test_pageserver_compaction_circuit_breaker: release-x86-64-with-lfc, release-arm64-with-lfc
• test_sharding_smoke: release-x86-64-with-lfc, release-arm64-with-lfc
• test_sharding_backpressure: release-x86-64-with-lfc, release-arm64-with-lfc

Failures on Postgres 14

• test_sharding_smoke: release-x86-64-with-lfc, release-arm64-with-lfc
• test_sharding_backpressure: release-x86-64-with-lfc, release-arm64-with-lfc

# Run all failed tests locally:
scripts/pytest -vv -n $(nproc) -k "test_sharding_smoke[release-pg14] or test_sharding_smoke[release-pg14] or test_sharding_backpressure[release-pg14] or test_sharding_backpressure[release-pg14] or test_pageserver_compaction_circuit_breaker[release-pg15] or test_pageserver_compaction_circuit_breaker[release-pg15] or test_sharding_smoke[release-pg15] or test_sharding_smoke[release-pg15] or test_sharding_backpressure[release-pg15] or test_sharding_backpressure[release-pg15] or test_pageserver_compaction_circuit_breaker[release-pg16] or test_pageserver_compaction_circuit_breaker[release-pg16] or test_sharding_smoke[release-pg16] or test_sharding_smoke[release-pg16] or test_sharding_backpressure[release-pg16] or test_sharding_backpressure[release-pg16] or test_pageserver_compaction_circuit_breaker[release-pg17] or test_pageserver_compaction_circuit_breaker[release-pg17] or test_pageserver_compaction_circuit_breaker[release-pg17] or test_pageserver_compaction_circuit_breaker[release-pg17] or test_sharding_backpressure[release-pg17] or test_sharding_backpressure[release-pg17] or test_sharding_backpressure[debug-pg17] or test_sharding_backpressure[release-pg17] or test_sharding_backpressure[release-pg17] or test_sharding_gc[release-pg17] or test_sharding_gc[release-pg17] or test_sharding_gc[release-pg17] or test_sharding_gc[release-pg17] or test_sharding_smoke[debug-pg17] or test_sharding_smoke[release-pg17] or test_sharding_smoke[release-pg17]"

Flaky tests (3)

Postgres 17

• test_sharding_smoke: release-x86-64-with-lfc
• test_s3_wal_replay: release-x86-64-without-lfc

Postgres 14

• test_pageserver_gc_compaction_idempotent[after_restart]: release-x86-64-with-lfc

Test coverage report is not available

_{The comment gets automatically updated with the latest test results
ccbb85d at 2025-02-06T17:03:28.727Z :recycle:}

[ololobus]

I am not sure why so many sharding tests failed. If the list of pageserver changes, the endpoint should still be reconfigured. It looks like there were some implicit assumptions that the endpoint is reconfigured with catalog updates. I had a brief look, still not clear. Will continue tomorrow