12 KiB
12 KiB
Unified Startup Sequence Design
Overview
This document describes the new unified startup sequence where all config values are created first, then CLI overrides are applied as a separate atomic operation. This eliminates the current 3-step incremental building process.
Current Problems
-
Incremental Config Building: Config is built in 3 steps:
- Step 1:
populate_default_config_values()- adds defaults - Step 2: CLI overrides applied via
update_config_in_table() - Step 3:
add_pubkeys_to_config_table()- adds generated keys
- Step 1:
-
Race Conditions: Cache can be refreshed between steps, causing incomplete config reads
-
Complexity: Multiple code paths for first-time vs restart scenarios
New Design Principles
- Atomic Config Creation: All config values created in single transaction
- Separate Override Phase: CLI overrides applied after complete config exists
- Unified Code Path: Same logic for first-time and restart scenarios
- Cache Safety: Cache only loaded after config is complete
Scenario 1: First-Time Startup (No Database)
Sequence
1. Key Generation Phase
├─ generate_random_private_key_bytes() → admin_privkey_bytes
├─ nostr_bytes_to_hex() → admin_privkey (hex)
├─ nostr_ec_public_key_from_private_key() → admin_pubkey_bytes
├─ nostr_bytes_to_hex() → admin_pubkey (hex)
├─ generate_random_private_key_bytes() → relay_privkey_bytes
├─ nostr_bytes_to_hex() → relay_privkey (hex)
├─ nostr_ec_public_key_from_private_key() → relay_pubkey_bytes
└─ nostr_bytes_to_hex() → relay_pubkey (hex)
2. Database Creation Phase
├─ create_database_with_relay_pubkey(relay_pubkey)
│ └─ Sets g_database_path = "<relay_pubkey>.db"
└─ init_database(g_database_path)
└─ Creates database with embedded schema (includes config table)
3. Complete Config Population Phase (ATOMIC)
├─ BEGIN TRANSACTION
├─ populate_all_config_values_atomic()
│ ├─ Insert ALL default config values from DEFAULT_CONFIG_VALUES[]
│ ├─ Insert admin_pubkey
│ └─ Insert relay_pubkey
└─ COMMIT TRANSACTION
4. CLI Override Phase (ATOMIC)
├─ BEGIN TRANSACTION
├─ apply_cli_overrides()
│ ├─ IF cli_options.port_override > 0:
│ │ └─ UPDATE config SET value = ? WHERE key = 'relay_port'
│ ├─ IF cli_options.admin_pubkey_override[0]:
│ │ └─ UPDATE config SET value = ? WHERE key = 'admin_pubkey'
│ └─ IF cli_options.relay_privkey_override[0]:
│ └─ UPDATE config SET value = ? WHERE key = 'relay_privkey'
└─ COMMIT TRANSACTION
5. Secure Key Storage Phase
└─ store_relay_private_key(relay_privkey)
└─ INSERT INTO relay_seckey (private_key_hex) VALUES (?)
6. Cache Initialization Phase
└─ refresh_unified_cache_from_table()
└─ Loads complete config into g_unified_cache
Function Call Sequence
// In main.c - first_time_startup branch
if (is_first_time_startup()) {
// 1. Key Generation
first_time_startup_sequence(&cli_options);
// → Generates keys, stores in g_unified_cache
// → Sets g_database_path
// → Does NOT populate config yet
// 2. Database Creation
init_database(g_database_path);
// → Creates database with schema
// 3. Complete Config Population (NEW FUNCTION)
populate_all_config_values_atomic(&cli_options);
// → Inserts ALL defaults + pubkeys in single transaction
// → Does NOT apply CLI overrides yet
// 4. CLI Override Phase (NEW FUNCTION)
apply_cli_overrides_atomic(&cli_options);
// → Updates config table with CLI overrides
// → Separate transaction after complete config exists
// 5. Secure Key Storage
store_relay_private_key(relay_privkey);
// 6. Cache Initialization
refresh_unified_cache_from_table();
}
New Functions Needed
// In config.c
int populate_all_config_values_atomic(const cli_options_t* cli_options) {
// BEGIN TRANSACTION
// Insert ALL defaults from DEFAULT_CONFIG_VALUES[]
// Insert admin_pubkey from g_unified_cache
// Insert relay_pubkey from g_unified_cache
// COMMIT TRANSACTION
return 0;
}
int apply_cli_overrides_atomic(const cli_options_t* cli_options) {
// BEGIN TRANSACTION
// IF port_override: UPDATE config SET value = ? WHERE key = 'relay_port'
// IF admin_pubkey_override: UPDATE config SET value = ? WHERE key = 'admin_pubkey'
// IF relay_privkey_override: UPDATE config SET value = ? WHERE key = 'relay_privkey'
// COMMIT TRANSACTION
// invalidate_config_cache()
return 0;
}
Scenario 2: Restart with Existing Database + CLI Options
Sequence
1. Database Discovery Phase
├─ find_existing_db_files() → ["<relay_pubkey>.db"]
├─ extract_pubkey_from_filename() → relay_pubkey
└─ Sets g_database_path = "<relay_pubkey>.db"
2. Database Initialization Phase
└─ init_database(g_database_path)
└─ Opens existing database
3. Config Validation Phase
└─ validate_config_table_completeness()
├─ Check if all required keys exist
└─ IF missing keys: populate_missing_config_values()
4. CLI Override Phase (ATOMIC)
├─ BEGIN TRANSACTION
├─ apply_cli_overrides()
│ └─ UPDATE config SET value = ? WHERE key = ?
└─ COMMIT TRANSACTION
5. Cache Initialization Phase
└─ refresh_unified_cache_from_table()
└─ Loads complete config into g_unified_cache
Function Call Sequence
// In main.c - existing relay branch
else {
// 1. Database Discovery
char** existing_files = find_existing_db_files();
char* relay_pubkey = extract_pubkey_from_filename(existing_files[0]);
startup_existing_relay(relay_pubkey);
// → Sets g_database_path
// 2. Database Initialization
init_database(g_database_path);
// 3. Config Validation (NEW FUNCTION)
validate_config_table_completeness();
// → Checks for missing keys
// → Populates any missing defaults
// 4. CLI Override Phase (REUSE FUNCTION)
if (has_cli_overrides(&cli_options)) {
apply_cli_overrides_atomic(&cli_options);
}
// 5. Cache Initialization
refresh_unified_cache_from_table();
}
New Functions Needed
// In config.c
int validate_config_table_completeness(void) {
// Check if all DEFAULT_CONFIG_VALUES keys exist
// IF missing: populate_missing_config_values()
return 0;
}
int populate_missing_config_values(void) {
// BEGIN TRANSACTION
// For each key in DEFAULT_CONFIG_VALUES:
// IF NOT EXISTS: INSERT INTO config
// COMMIT TRANSACTION
return 0;
}
int has_cli_overrides(const cli_options_t* cli_options) {
return (cli_options->port_override > 0 ||
cli_options->admin_pubkey_override[0] != '\0' ||
cli_options->relay_privkey_override[0] != '\0');
}
Scenario 3: Restart with Existing Database + No CLI Options
Sequence
1. Database Discovery Phase
├─ find_existing_db_files() → ["<relay_pubkey>.db"]
├─ extract_pubkey_from_filename() → relay_pubkey
└─ Sets g_database_path = "<relay_pubkey>.db"
2. Database Initialization Phase
└─ init_database(g_database_path)
└─ Opens existing database
3. Config Validation Phase
└─ validate_config_table_completeness()
├─ Check if all required keys exist
└─ IF missing keys: populate_missing_config_values()
4. Cache Initialization Phase (IMMEDIATE)
└─ refresh_unified_cache_from_table()
└─ Loads complete config into g_unified_cache
Function Call Sequence
// In main.c - existing relay branch (no CLI overrides)
else {
// 1. Database Discovery
char** existing_files = find_existing_db_files();
char* relay_pubkey = extract_pubkey_from_filename(existing_files[0]);
startup_existing_relay(relay_pubkey);
// 2. Database Initialization
init_database(g_database_path);
// 3. Config Validation
validate_config_table_completeness();
// 4. Cache Initialization (IMMEDIATE - no overrides to apply)
refresh_unified_cache_from_table();
}
Key Improvements
1. Atomic Config Creation
Before:
populate_default_config_values(); // Step 1
update_config_in_table("relay_port", port_str); // Step 2
add_pubkeys_to_config_table(); // Step 3
After:
populate_all_config_values_atomic(&cli_options); // Single transaction
apply_cli_overrides_atomic(&cli_options); // Separate transaction
2. Elimination of Race Conditions
Before:
- Cache could refresh between steps 1-3
- Incomplete config could be read
After:
- Config created atomically
- Cache only refreshed after complete config exists
3. Unified Code Path
Before:
- Different logic for first-time vs restart
populate_default_config_values()vsadd_pubkeys_to_config_table()
After:
- Same validation logic for both scenarios
validate_config_table_completeness()handles both cases
4. Clear Separation of Concerns
Before:
- CLI overrides mixed with default population
- Unclear when overrides are applied
After:
- Phase 1: Complete config creation
- Phase 2: CLI overrides (if any)
- Phase 3: Cache initialization
Implementation Changes Required
1. New Functions in config.c
// Atomic config population for first-time startup
int populate_all_config_values_atomic(const cli_options_t* cli_options);
// Atomic CLI override application
int apply_cli_overrides_atomic(const cli_options_t* cli_options);
// Config validation for existing databases
int validate_config_table_completeness(void);
int populate_missing_config_values(void);
// Helper function
int has_cli_overrides(const cli_options_t* cli_options);
2. Modified Functions in config.c
// Simplify to only generate keys and set database path
int first_time_startup_sequence(const cli_options_t* cli_options);
// Remove config population logic
int add_pubkeys_to_config_table(void); // DEPRECATED - logic moved to populate_all_config_values_atomic()
3. Modified Startup Flow in main.c
// First-time startup
if (is_first_time_startup()) {
first_time_startup_sequence(&cli_options);
init_database(g_database_path);
populate_all_config_values_atomic(&cli_options); // NEW
apply_cli_overrides_atomic(&cli_options); // NEW
store_relay_private_key(relay_privkey);
refresh_unified_cache_from_table();
}
// Existing relay
else {
startup_existing_relay(relay_pubkey);
init_database(g_database_path);
validate_config_table_completeness(); // NEW
if (has_cli_overrides(&cli_options)) {
apply_cli_overrides_atomic(&cli_options); // NEW
}
refresh_unified_cache_from_table();
}
Benefits
- Atomicity: Config creation is atomic - no partial states
- Simplicity: Clear phases with single responsibility
- Safety: Cache only loaded after complete config exists
- Consistency: Same validation logic for all scenarios
- Maintainability: Easier to understand and modify
- Testability: Each phase can be tested independently
Migration Path
- Implement new functions in config.c
- Update main.c startup flow
- Test first-time startup scenario
- Test restart with CLI overrides
- Test restart without CLI overrides
- Remove deprecated functions
- Update documentation
Testing Strategy
Test Cases
-
First-time startup with defaults
- Verify all config values created atomically
- Verify cache loads complete config
-
First-time startup with port override
- Verify defaults created first
- Verify port override applied second
- Verify cache reflects override
-
Restart with complete config
- Verify no config changes
- Verify cache loads immediately
-
Restart with missing config keys
- Verify missing keys populated
- Verify existing keys unchanged
-
Restart with CLI overrides
- Verify overrides applied atomically
- Verify cache invalidated and refreshed
Validation Points
- Config table row count after each phase
- Cache validity state after each phase
- Transaction boundaries (BEGIN/COMMIT)
- Error handling for failed transactions