feat: add --alter-logs-indexes-retention-days flag

[michael-richey]

Summary

Add a new sync/migrate flag, --alter-logs-indexes-retention-days, that overrides the standard (non-flex) retention window on logs indexes. It mirrors the existing --alter-flex-logs-retention-days flag, which overrides the flex-tier retention window.

Motivation

--alter-flex-logs-retention-days already lets operators override num_flex_logs_retention_days on synced logs indexes. There was no equivalent for the standard retention field, num_retention_days. This adds the symmetric flag so operators can adjust non-flex retention during sync/migrate without editing imported state by hand.

Behavior

• New flag: --alter-logs-indexes-retention-days, type=IntRange(min=30), default=None.
• Registered in _sync_options, so it is available on both sync and migrate (same as the flex flag).
• In LogsIndexes.pre_resource_action_hook, when the flag is set and num_retention_days is already present on the resource, the field is overwritten with the supplied value. Insert-if-present semantics: a resource that lacks the field is never modified — the field is never added, only overridden. This matches the flex flag exactly.
• The two retention flags are independent: setting one does not affect the other's field.

Implementation

pre_resource_action_hook previously had a single hardcoded block for the flex field. Rather than duplicating it for the new field, the hook now drives both overrides from a class-level table:

_RETENTION_OVERRIDES = (
    ("alter_flex_logs_retention_days", "num_flex_logs_retention_days"),
    ("alter_logs_indexes_retention_days", "num_retention_days"),
)

async def pre_resource_action_hook(self, _id, resource: Dict) -> None:
    for config_attr, field in self._RETENTION_OVERRIDES:
        retention_days = getattr(self.config, config_attr)
        if retention_days is not None and field in resource:
            resource[field] = retention_days

The flex behavior is byte-for-byte identical to before; future retention overrides are now a one-line table addition.

Config plumbing (Configuration dataclass field, build_config kwargs extraction, constructor kwarg) follows the same pattern as the flex flag.

Tests

New file tests/unit/test_logs_indexes_retention.py (mirrors test_logs_indexes_flex_retention.py):

• CLI acceptance on sync and migrate (flag value reaches run_cmd as alter_logs_indexes_retention_days).
• CLI rejection of invalid values (invalid, -1, 0, 29 — below the min=30 floor).
• Hook overrides an existing num_retention_days.
• Hook does not add num_retention_days to a resource lacking it.
• Unset flag (None) preserves the existing num_retention_days.
• Regression: both retention flags apply independently.
• Regression: the new flag does not touch the flex field.

All values are synthetic (index-test, integer day counts).

Test plan

• tox -e py311 -- tests/unit/test_logs_indexes_retention.py tests/unit/test_logs_indexes_flex_retention.py — 20 passed
• tox -e ruff — clean
• Full unit suite (pytest tests/unit) — no new failures vs. origin/main (9 pre-existing subprocess-test failures are unrelated and reproduce on clean main)

[michael-richey]

Reviewed this end-to-end — nice addition. The new --alter-logs-indexes-retention-days flag is wired cleanly and mirrors the existing flex-retention override behavior with the same insert-if-present semantics.

What I validated:

• CLI option registration in datadog_sync/commands/shared/options.py.
• Config plumbing through Configuration and build_config in datadog_sync/utils/configuration.py.
• Hook behavior in datadog_sync/model/logs_indexes.py after the table-driven refactor.
• New unit tests in tests/unit/test_logs_indexes_retention.py (including independence between the two retention flags and no-field insertion guard).
• Local test run: tests/unit/test_logs_indexes_retention.py + tests/unit/test_logs_indexes_flex_retention.py (20 passed).

Non-blocking suggestion: add a migrate-path invalid-value case (parallel to the sync invalid-value test), since this option is shared across both commands.

[michael-richey]

Completed a code + test review; details captured in the PR comment. No blocking issues found.

[michael-richey]

Thanks for the review. Addressed the non-blocking suggestion in 12d4fdb: test_cli_rejects_invalid_logs_indexes_retention_days is now parametrized across both sync and migrate (mirroring test_cli_accepts_alter_logs_indexes_retention_days), so the shared --alter-logs-indexes-retention-days option is rejected for invalid values on both commands. Local run: tests/unit/test_logs_indexes_retention.py + tests/unit/test_logs_indexes_flex_retention.py — 24 passed; ruff clean.