2024-09-05 11:23:15 +00:00
# NeoFS block storage
Using NeoFS to store chain's blocks and snapshots was proposed in
[#3463 ](https://github.com/neo-project/neo/issues/3463 ). NeoGo contains several
extensions utilizing NeoFS block storage aimed to improve node synchronization
efficiency and reduce node storage size.
## Components and functionality
### Block storage schema
2024-10-21 20:49:09 +00:00
A single NeoFS container is used to store blocks and index files. Each container
has network magic attribute (`Magic:56753`). Each block is stored in a binary
form as a separate object with a unique OID and a set of attributes:
- block object identifier with block index value (`Block:1`)
- primary node index (`Primary:0`)
- block hash in the LE form (`Hash:5412a781caf278c0736556c0e544c7cfdbb6e3c62ae221ef53646be89364566b`)
- previous block hash in the LE form (`PrevHash:3654a054d82a8178c7dfacecc2c57282e23468a42ee407f14506368afe22d929`)
2024-12-12 17:03:24 +00:00
- millisecond-precision block creation timestamp (`BlockTime:1627894840919`)
- second-precision block uploading timestamp (`Timestamp:1627894840`)
2024-09-05 11:23:15 +00:00
Each index file is an object containing a constant-sized batch of raw block object
IDs in binary form ordered by block index. Each index file is marked with the
following attributes:
2024-10-21 20:49:09 +00:00
- index file identifier with consecutive file index value (`Index:0`)
- the number of OIDs included into index file (`IndexSize:128000`)
2024-12-12 17:03:24 +00:00
- second-precision index file uploading timestamp (`Timestamp:1627894840`)
2024-09-05 11:23:15 +00:00
### NeoFS BlockFetcher
NeoFS BlockFetcher service is designed as an alternative to P2P synchronisation
protocol. It allows to download blocks from a trusted container in the NeoFS network
and persist them to database using standard verification flow. NeoFS BlockFetcher
service primarily used during the node's bootstrap, providing a fast alternative to
P2P blocks synchronisation.
NeoFS BlockFetcher service has two modes of operation:
- Index File Search: Search for index files, which contain batches of block object
IDs and fetch blocks from NeoFS by retrieved OIDs.
- Direct Block Search: Search and fetch blocks directly from NeoFS container via
built-in NeoFS object search mechanism.
Operation mode of BlockFetcher can be configured via `SkipIndexFilesSearch`
parameter.
#### Operation flow
1. **OID Fetching** :
Depending on the mode, the service either:
- Searches for index files by index file attribute and reads block OIDs from index
file object-by-object.
2024-11-21 14:56:55 +00:00
- Searches blocks one by one directly by block attribute.
2024-09-05 11:23:15 +00:00
Once the OIDs are retrieved, they are immediately redirected to the
block downloading routines for further processing. The channel that
is used to redirect block OIDs to downloading routines is buffered
to provide smooth OIDs delivery without delays. The size of this channel
can be configured via `OIDBatchSize` parameter and equals to `2*OIDBatchSize` .
2. **Parallel Block Downloading** :
The number of downloading routines can be configured via
`DownloaderWorkersCount` parameter. It's up to the user to find the
balance between the downloading speed and blocks persist speed for every
blockfetcher: fix block enqueue logic
Previously, the `blockQueuer` routine, which enqueues blocks into
`bQueue`, could be blocked on enqueing newer blocks if older blocks
downloading is delayed by NeoFS.
The `blocksCh` channel, acting as a queue ordered by download speed,
conflicted with the BQueue requirement for strict sequential enqueuing
(expecting an exact range of blocks), resulting in a deadlock that
stalled the process.
Before with default config settings:
```
2024-11-27T17:12:19.348+0300 INFO persisted to disk {"blocks":
0, "keys": 116, "headerHeight": 0, "blockHeight": 0, "took": "15
.509083ms"}
2024-11-27T17:19:39.574+0300 INFO persisted to disk {"blocks":
16, "keys": 11107, "headerHeight": 216768, "blockHeight": 216768,
"took": "62.762041ms"}
```
Average block persistence speed: 492.40 block/s
Average blocks number for each persist log: 584.28
After:
```
2024-11-27T17:29:03.362+0300 INFO persisted to disk {"blocks":
0, "keys": 116, "headerHeight": 0, "blockHeight": 0, "took": "19
.485084ms"}
2024-11-27T17:34:58.527+0300 INFO persisted to disk {"blocks":
16, "keys": 11109, "headerHeight": 216770, "blockHeight": 216769,
"took": "52.43925ms"}
```
Average block persistence speed: 610.33 block/s
Average blocks number for each persist log: 752.61
Close #3699
Signed-off-by: Ekaterina Pavlova <ekt@morphbits.io>
2024-11-27 15:23:38 +00:00
node that uses NeoFS BlockFetcher. Downloaded blocks are placed to the
block queue directly.
2024-09-05 11:23:15 +00:00
3. **Block Insertion** :
Downloaded blocks are inserted into the blockchain using the same logic
as in the P2P synchronisation protocol. The block queue is used to order
downloaded blocks before they are inserted into the blockchain. The
size of the queue can be configured via the `BQueueSize` parameter
and should be larger than the `OIDBatchSize` parameter to avoid blocking
the downloading routines.
Once all blocks available in the NeoFS container are processed, the service
shuts down automatically.
2024-10-15 12:27:28 +00:00
### NeoFS Upload Command
The `upload-bin` command is designed to fetch blocks from the RPC node and upload
them to the NeoFS container.
It also creates and uploads index files. Below is an example usage of the command:
```shell
2024-10-21 20:49:09 +00:00
./bin/neo-go util upload-bin --cid 9iVfUg8aDHKjPC4LhQXEkVUM4HDkR7UCXYLs8NQwYfSG --wallet-config ./wallet-config.yml --block-attribute Block --index-attribute Index --rpc-endpoint https://rpc.t5.n3.nspcc.ru:20331 -fsr st1.t5.fs.neo.org:8080 -fsr st2.t5.fs.neo.org:8080 -fsr st3.t5.fs.neo.org:8080
2024-10-15 12:27:28 +00:00
```
The command supports the following options:
```
NAME:
neo-go util upload-bin - Fetch blocks from RPC node and upload them to the NeoFS container
USAGE:
2024-12-07 14:59:05 +00:00
neo-go util upload-bin --fs-rpc-endpoint < address1 > [,< address2 > [...]] --container < cid > --block-attribute block --index-attribute index --rpc-endpoint < node > [--timeout < time > ] --wallet < wallet > [--wallet-config < config > ] [--address < address > ] [--workers < num > ] [--searchers < num > ] [--index-file-size < size > ] [--retries < num > ] [--debug]
2024-10-15 12:27:28 +00:00
OPTIONS:
--fs-rpc-endpoint value, --fsr value [ --fs-rpc-endpoint value, --fsr value ] List of NeoFS storage node RPC addresses (comma-separated or multiple --fs-rpc-endpoint flags)
--container value, --cid value NeoFS container ID to upload blocks to
--block-attribute value Attribute key of the block object
--index-attribute value Attribute key of the index file object
--address value Address to use for signing the uploading and searching transactions in NeoFS
--index-file-size value Size of index file (default: 128000)
2024-10-23 18:41:41 +00:00
--workers value Number of workers to fetch, upload and search blocks concurrently (default: 50)
--searchers value Number of concurrent searches for blocks (default: 20)
2024-11-02 18:56:06 +00:00
--retries value Maximum number of Neo/NeoFS node request retries (default: 5)
2024-11-15 12:35:37 +00:00
--debug, -d Enable debug logging (LOTS of output, overrides configuration) (default: false)
2024-10-15 12:27:28 +00:00
--rpc-endpoint value, -r value RPC node address
--timeout value, -s value Timeout for the operation (default: 10s)
--wallet value, -w value Wallet to use to get the key for transaction signing; conflicts with --wallet-config flag
--wallet-config value Path to wallet config to use to get the key for transaction signing; conflicts with --wallet flag
--help, -h show help
```
This command works as follows:
1. Fetches the current block height from the RPC node.
2024-12-07 14:59:05 +00:00
2. Searches for the index files stored in NeoFS.
3. Searches for the stored blocks from the latest incomplete index file.
4. Fetches missing blocks from the RPC node and uploads them to the NeoFS container.
5. After uploading the blocks, it creates index file based on the uploaded block OIDs.
6. Uploads the created index file to the NeoFS container.
7. Repeats steps 4-6 until the current block height is reached.
2024-10-15 12:27:28 +00:00
If the command is interrupted, it can be resumed. It starts the uploading process
2024-12-07 14:59:05 +00:00
from the last uploaded index file.
For a given block sequence, only one type of index file is supported. If new index
files are needed (different `index-file-size` or `index-attribute` ), the entire
block sequence must be uploaded from the beginning. Please, add a comment to the
[#3744 ](https://github.com/nspcc-dev/neo-go/issues/3744 ) issue if you need this
functionality.