* Accelerator: partial fullnodes
1) Node can monitor a subset of shards
2) New archive slice format (sharded)
3) Validators are still required to have all shards
4) Support partial liteservers in lite-client, blockchain explorer, tonlib
5) Proxy liteserver
* Fix compilation error
1) Remove config 41, move "full collated data" to capabilities
2) Whitelist on collator nodes
3) "Ping" request for collator nodes
4) More customizable collators list for validators
5) CollationManager
* Persistent state download improvements
1) Don't start over on restart
2) Download shards one at a time to reduce RAM usage
3) More logs
* Remove old peers from adnl stats
* getactorstats query for validator-engine-console
* celldb in-memory mode (--celldb-in-memory option)
* rldp2: bugfix - do not estimate speed while nothing is sent
* add simple ed25519 benchmark
* fix compilation errors of different platforms and move to c++20
* fix some warnings
* turn on TON_USE_ABSEIL for glibc 2.27 nix build
---------
Co-authored-by: birydrad <>
* Validator improvements
* Fix cancelling state serialization
* Disable state serializer on all mainnet validators
* Flag --catchain-max-block-delay-slow
* Set default catchain-max-block-delay to 0.4, delay-slow to 1.0
---------
Co-authored-by: SpyCheese <mikle98@yandex.ru>
* Add --fast-state-serializer and tools for jemalloc
* Disable fast state serializer by default unless RAM is >= 90GB
* Print jemalloc stats once a minute
* Dump jemalloc profile on request
* Respond to AdnlMessageCreateChannel with Nop
---------
Co-authored-by: SpyCheese <mikle98@yandex.ru>
* Set collator options from validator console
* Fix compilation error in manager-disk
* Defer all messages if out msg queue is too big
* Fix checking queue size in collator
---------
Co-authored-by: SpyCheese <mikle98@yandex.ru>
* Deferred messages and msg metadata
* Store out msg queue size in state
* Add checks for queue processing
1. Collator must process at least one message from AccountDispatchQueue (unless block is full)
2. The first message from a transaction is not counted, it cannot be deferred (unless AccountDispatchQueue is not empty)
* Return msg metadata from LS in listBlockTransactions[Ext]
* Enable new features by capabilities
* Changes in deferred messages
* Process deferred messages via new_msgs in collator
* Rework setting deferred_lt, bring back check_message_processing_order, check order of deferred_lt in validator
* Use have_unprocessed_account_dispatch_queue_ in collator
* Fix setting transaction lt for deferred messages
* Fix lite-client compilation error
* Changes in process_dispatch_queue, rename deferred_lt -> emitted_lt
* Fix compilation error
* Use uint64 for msg queue size
* Add liteServer.getBlockOutMsgQueueSize
* Fix compilation error
* Fix typos in comments
---------
Co-authored-by: SpyCheese <mikle98@yandex.ru>
* Private overlay for external messages
* Improve ext msg overlays
* Manage from validator console
* Bypass out queue size limit for high-priority messages
* Shuffle messages in get_external_messages
* Cleanup mempool when creating validator group
* Improve private overlays for externals
1. Allow using validator adnl ids in addition to fullnode ids
2. Set priority per sender, not per overlay
3. Require the same overlay name for all nodes
4. Enable lz4 in private block overlay
* Fix typo, add debug logs
* Enable lz4 in private block overlay by config
Change proto_version for lz4 in catchain overlays to 4
* Add logs for broadcasts in fullnode
---------
Co-authored-by: SpyCheese <mikle98@yandex.ru>
