[BorisTyshkevich]

Refs #692

Corrections (no change to the upload mechanism; that is part 2):

1. Stale sizes: removed frozen 65 KB / 500 KB / 230 KB figures; dist/sql.html is about 2 MB raw, explained raw vs compressed transfer and how to measure; option C now notes the 1 MiB ConfigMap cap.
2. Installer prerequisites: npm, node 22+, curl (OIDC discovery only), clickhouse-client; prebuilt release sql.html path mentioned.
3. Applicability: native hosting vs container runtime; providers that forbid http_handlers; option B replication scope (shard/replication group only); option C ConfigMap cap.
4. Auth wording: Bearer id_token by default, access_token with audience, JWT as Basic password with ch_auth basic; install example shows --ch-auth basic and --audience; http_handlers.xml comment and README alt text fixed.
5. install.sh printed https even without --secure and omitted the HTTP port; now a print_next_steps function (also shown by --dry-run) with explicit scheme and default 8123/8443; regression test tests/unit/install-script.test.js covers secure and non-secure.
6. .wiki/Deployment-and-Security.md: nginx -> Caddy.

CHANGELOG [Unreleased] entry added.

• Correction (d88cc4d): empirical Docker testing (25.8 OSS, 26.3 OSS, Antalya 25.8.33) showed install.sh --cluster fails on any multi-replica cluster (STORAGE_REQUIRES_PARAMETER) and leaves a temp table behind. Docs/comments no longer claim fan-out; added a tested interim ON CLUSTER ... File ... AS SELECT FROM url() procedure to docs/ASSET-DISTRIBUTION.md (release-URL pull and sql-config.json untested). Installer fix is PR 2.

• install.sh --cluster guard now fails closed (exit 2, before building) unless the system.clusters count query succeeds and prints exactly 1; distinct messages for query failure, unknown cluster (0), malformed output, and more than one replica (pointing at the manual procedure). The manual multi-node procedure is now an idempotent pre-clean/create/drop/verify sequence (retry flow untested). Docs also gain the sql-config.json cluster step (tested single-node 26.3 only) and cleanup DROP ... ON CLUSTER for tables left by earlier failed runs.

🤖 Generated with Claude Code

https://claude.ai/code/session_01P3CopPwrpzeWAJd7vaaLSk