feat(logging): always-on startup logging

[EddieHouston]

Summary

• Electrs has been seen dying silently during startup in environments
  running with no -v flag (error-only logging), making it impossible to
  diagnose where they got stuck
• Logging is now initialized at info level minimum during startup, then
  drops to the user's configured verbosity after servers are listening
• Adds info! milestone logs at each startup phase: daemon connection,
  DB open, initial sync start/complete, mempool load, startup complete
• Promotes key debug! logs in schema.rs and mempool.rs to info!
  so block/header/mempool counts are visible during the startup window

Test plan

• Run electrs with no -v flag — verify startup milestones are
  logged and logging goes silent after "startup complete"
• Run electrs with -vvv — verify behavior is unchanged (already
  above the info threshold)
• Kill electrs mid-startup — verify the last log line indicates
  which phase it was in

[philippem]

let's wait until we see

2026-04-14T11:28:46.020-07:00 - INFO Electrum RPC server running on 0.0.0.0:50001
2026-04-14T11:28:46.021-07:00 - INFO REST server running on 0.0.0.0:3001

before we revert verbosity. it's only at this point that electrs is ready

• print the previous and after levels
• the "Config" print message at startup should show the actual requested log level, not the overridden one:

$ ./run-release.sh
Config { log: StdErrLog { verbosity: Info, quiet: false, show_level: true, timestamp: Millisecond, modules: [], writer: "stderr", color_choice: Never, show_module_names: false }, network_type: Bitcoin, db_path: "/mnt/quattro/electrs/bitcoin-mainnet/mainnet", daemon_dir: "/home/phil/.bitcoin", blocks_dir: "/home/phil/.bitcoin/blocks", daemon_rpc_addr: 0.0.0.0:8332, daemon_parallelism: 8, cookie: Some("rpcuser:rpcpassword"), electrum_rpc_addr: 0.0.0.0:50001, http_addr: 0.0.0.0:3001, http_socket_file: None, monitoring_addr: 0.0.0.0:4224, jsonrpc_import: true, light_mode: false, address_search: false, index_unspendables: false, cors: Some("*"), precache_scripts: None, utxos_limit: 500, electrum_txs_limit: 500, electrum_banner: "Welcome to electrs-esplora 0.4.1", rpc_logging: RpcLogging { enabled: false, hide_params: false, anonymize_ip: false }, zmq_addr: None, log_verbosity: 1, db_block_cache_mb: 8192, db_parallelism: 20, db_write_buffer_size_mb: 256, initial_sync_batch_size: 100, db_cache_index_filter_blocks: false }

this is what it should show:

$ ./run-release.sh
Config { log: StdErrLog { verbosity: Error, quiet: false, show_level: true, timestamp: Millisecond, modules: [], writer: "stderr", color_choice: Auto, show_module_names: false }, network_type: Bitcoin, db_path: "/mnt/quattro/electrs/bitcoin-mainnet/mainnet", daemon_dir: "/home/phil/.bitcoin", blocks_dir: "/home/phil/.bitcoin/blocks", daemon_rpc_addr: 0.0.0.0:8332, daemon_parallelism: 8, cookie: Some("rpcuser:rpcpassword"), electrum_rpc_addr: 0.0.0.0:50001, http_addr: 0.0.0.0:3001, http_socket_file: None, monitoring_addr: 0.0.0.0:4224, jsonrpc_import: true, light_mode: false, address_search: false, index_unspendables: false, cors: Some("*"), precache_scripts: None, utxos_limit: 500, electrum_txs_limit: 500, electrum_banner: "Welcome to electrs-esplora 0.4.1", rpc_logging: RpcLogging { enabled: false, hide_params: false, anonymize_ip: false }, zmq_addr: None, db_block_cache_mb: 8192, db_parallelism: 20, db_write_buffer_size_mb: 256, initial_sync_batch_size: 100, db_cache_index_filter_blocks: false }

• another way to implement this would be to add a second option for startup logging. that would simplify the set/revert mechanics

[shesek]

I would consider a different approach: make level 2 logging (error+warn+info) the default and the minimum, such that a single -v enables debug-level logging and -vv enables trace (i.e. log.verbosity(2 + m.occurrences_of("verbosity"))). Then promote important log messages to be info-level (this PR already promotes some messages, but possibly too many. I think some should remain debug).

I don't think it's really ever desirable to hide error/warn/info messages, and it's typical for CLI programs to output these by default without requiring -v, so that behavior makes more sense to me. And it's also simpler than the revert mechanism introduced here.

Changing the meaning of -v does break backwards compatibility, but I think it's something we could live with. The (probably?) most typical setting, -vvvv, will still work the same. If we want to continue supporting logging level < 2, we could introduce a separate -q/--quiet parameter which is the more typical way to handle this, but I don't think that's really necessary.

[EddieHouston]

I would consider a different approach: make level 2 logging (error+warn+info) the default and the minimum, such that a single -v enables debug-level logging and -vv enables trace (i.e. log.verbosity(2 + m.occurrences_of("verbosity"))). Then promote important log messages to be info-level (this PR already promotes some messages, but possibly too many. I think some should remain debug).

I don't think it's really ever desirable to hide error/warn/info messages, and it's typical for CLI programs to output these by default without requiring -v, so that behavior makes more sense to me. And it's also simpler than the revert mechanism introduced here.

Changing the meaning of -v does break backwards compatibility, but I think it's something we could live with. The (probably?) most typical setting, -vvvv, will still work the same. If we want to continue supporting logging level < 2, we could introduce a separate -q/--quiet parameter which is the more typical way to handle this, but I don't think that's really necessary.

Good simplification. This does remove the need to override log levels. I can see demoting the mempool message changed in this PR back to debug. I could also see some other info logs we can demote to debug as well:

File	Line	Message	Why
src/rest.rs	688	handle {:?} {:?}	Every HTTP request — biggest noise source
src/electrum/server.rs	873	connected peer	Every peer connection
src/electrum/server.rs	887	disconnected peer	Every peer disconnection
src/daemon.rs	792	downloaded N/M block headers	Chatty during initial sync (bracket messages stay info)
src/electrum/discovery.rs	591	Full server list JSON dump	Large payload, periodic

If we later find that there is a need to implement the -q/--quiet option (depending on memory growth with 2 as the default)... we can then look at if an override on startup is necessary when that is set.