feat(satp-hermes): add crash recovery & rollback protocol

1. Implemented recovery & rollback using RPC-based message handlers. 2. Added rollback strategies for all SATP stages. 3. Integrated database log management for recovery and rollback. 4. Added cron jobs for scheduled crash detection and recovery initiation. Co-authored-by: Rafael Belchior <rafael.belchior@tecnico.ulisboa.pt> Co-authored-by: Carlos Amaro <carlosrscamaro@tecnico.ulisboa.pt> Signed-off-by: Yogesh01000100 <yogeshone678@gmail.com> chore(satp-hermes): improve DB management Signed-off-by: Rafael Belchior <rafael.belchior@tecnico.ulisboa.pt> chore(satp-hermes): crash recovery architecture Signed-off-by: Rafael Belchior <rafael.belchior@tecnico.ulisboa.pt> fix(recovery): enhance crash recovery and rollback implementation Signed-off-by: Yogesh01000100 <yogeshone678@gmail.com> refactor(recovery): consolidate logic and improve SATP message handling Signed-off-by: Yogesh01000100 <yogeshone678@gmail.com> feat(recovery): add rollback implementations Signed-off-by: Yogesh01000100 <yogeshone678@gmail.com> fix: correct return types and inits Signed-off-by: Yogesh01000100 <yogeshone678@gmail.com> fix: add unit tests and resolve rollbackstate Signed-off-by: Yogesh01000100 <yogeshone678@gmail.com> feat: add function processing logs from g2 Signed-off-by: Yogesh01000100 <yogeshone678@gmail.com> feat: add cron schedule for periodic crash checks Signed-off-by: Yogesh01000100 <yogeshone678@gmail.com> fix: resolve rollback condition and add tests Signed-off-by: Yogesh01000100 <yogeshone678@gmail.com> feat: add orchestrator communication layer using connect-RPC Signed-off-by: Yogesh01000100 <yogeshone678@gmail.com> feat: add rollback protocol rpc Signed-off-by: Yogesh01000100 <yogeshone678@gmail.com> fix: handle server log synchronization Signed-off-by: Yogesh01000100 <yogeshone678@gmail.com> fix: resolve gol errors, add unit tests Signed-off-by: Yogesh01000100 <yogeshone678@gmail.com> fix: handle server-side rollback Signed-off-by: Yogesh01000100 <yogeshone678@gmail.com> fix: resolve networkId in rollback strategies Signed-off-by: Yogesh01000100 <yogeshone678@gmail.com>
hyperledger-cacti · Jan 9, 2025 · 43367c9 · 43367c9
1 parent 109b7b4
commit 43367c9
Show file tree

Hide file tree

Showing 43 changed files with 6,478 additions and 57 deletions.
diff --git a/packages/cactus-plugin-satp-hermes/README.md b/packages/cactus-plugin-satp-hermes/README.md
@@ -57,6 +57,17 @@ The sequence diagram of SATP is pictured below.
 
 ![satp-sequence-diagram](https://i.imgur.com/SOdXFEt.png)
 
+### Crash Recovery Integration
+The crash recovery protocol ensures session consistency across all stages of SATP. Each session's state, logs, hashes, timestamps, and signatures are stored and recovered using the following mechanisms:
+
+1. **Session Logs**: A persistent log storage mechanism ensures crash-resilient state recovery.
+2. **Consistency Checks**: Ensures all messages and actions are consistent across both gateways and the connected ledgers.
+3. **Stage Recovery**: Recovers interrupted sessions by validating logs, hashes, timestamps, and signatures to maintain protocol integrity.
+4. **Rollback Operations**: In the event of a timeout or irrecoverable failure, rollback messages ensure the state reverts back the current stage.
+5. **Logging & Proofs**: The database is leveraged for state consistency and proof accountability across gateways.
+
+Refer to the [Crash Recovery Sequence](https://datatracker.ietf.org/doc/html/draft-belchior-satp-gateway-recovery) for more details.
+
 ### Application-to-Gateway API (API Type 1)
 We
 
@@ -76,17 +87,33 @@ There are Client and Server Endpoints for each type of message detailed in the S
   - CommitFinalV1Response
   - TransferCompleteV1Request
   - ClientV1Request
+### Crash Recovery Endpoints
+There are Client and Server gRPC Endpoints for the recovery and rollback messages:
 
-There are also defined the endpoints for the crash recovery procedure (there is still missing the endpoint to receive the Rollback mesage):
-  - RecoverV1Message
-  - RecoverUpdateV1Message
-  - RecoverUpdateAckV1Message
-  - RecoverSuccessV1Message
-  - RollbackV1Message
+- **Recovery Messages:**
+  - `RecoverV2Message`
+  - `RecoverV2SuccessMessage`
+  - `RecoverUpdateMessage`
+- **Rollback Messages:**
+  - `RollbackV2Message`
+  - `RollbackAckMessage`
 
 ## Use case
 Alice and Bob, in blockchains A and B, respectively, want to make a transfer of an asset from one to the other. Gateway A represents the gateway connected to Alice's blockchain. Gateway B represents the gateway connected to Bob's blockchain. Alice and Bob will run SATP, which will execute the transfer of the asset from blockchain A to blockchain B. The above endpoints will be called in sequence. Notice that the asset will first be locked on blockchain A and a proof is sent to the server-side. Afterward, the asset on the original blockchain is extinguished, followed by its regeneration on blockchain B.
 
+### Role of Crash Recovery in SATP
+In SATP, crash recovery ensures that asset transfers remain consistent and fault-tolerant across distributed ledgers. Key features include:
+- **Session Recovery**: Gateways synchronize state using recovery messages, ensuring continuity after failures.
+- **Rollback**: For irrecoverable errors, rollback procedures ensure safe reversion to previous states.
+- **Fault Resilience**: Enables recovery from crashes while maintaining the integrity of ongoing transfers.
+
+These features enhance reliability in scenarios where network or gateway disruptions occur during asset transfers.
+
+### Future Work
+
+- **Single-Gateway Topology Enhancement**  
+  The crash recovery and rollback mechanisms are implemented for configurations where client and server data are handled separately. For single-gateway setups, where both client and server data coexist in session, the current implementation of fetching a single log may not suffice. This requires to fetch multiple logs (X logs) `recoverSessions()` to differentiate and handle client and server-specific data accurately, to reconstruct the session back after the crash.
+
 ## Running the tests
 
 [A test of the entire protocol with manual calls to the methods, i.e. without ledger connectors and Open API.](https://github.com/hyperledger/cactus/blob/2e94ef8d3b34449c7b4d48e37d81245851477a3e/packages/cactus-plugin-satp-hermes/src/test/typescript/integration/satp.test.ts)
@@ -109,6 +136,16 @@ Alice and Bob, in blockchains A and B, respectively, want to make a transfer of
 
 [A test with a backup gateway resuming the protocol after the client gateway crashed.](https://github.com/hyperledger/cactus/tree/main/packages/cactus-plugin-satp-hermes/src/test/typescript/integration/backup-gateway-after-client-crash.test.ts)
 
+
+### Crash Recovery Tests
+- [Stage 1 Recovery Test](src/test/typescript/integration/recovery/recovery-stage-1.test.ts)
+- [Stage 2 Recovery Test](src/test/typescript/integration/recovery/recovery-stage-2.test.ts)
+- [Stage 3 Recovery Test](src/test/typescript/integration/recovery/recovery-stage-3.test.ts)
+- [Stage 0 Rollback Test](src/test/typescript/integration/rollback/rollback-stage-0.test.ts)
+- [Stage 1 Rollback Test](src/test/typescript/integration/rollback/rollback-stage-1.test.ts)
+- [Stage 2 Rollback Test](src/test/typescript/integration/rollback/rollback-stage-2.test.ts)
+- [Stage 3 Rollback Test](src/test/typescript/integration/rollback/rollback-stage-3.test.ts)
+
 For developers that want to test separate steps/phases of the SATP protocol, please refer to [these](https://github.com/hyperledger/cactus/blob/2e94ef8d3b34449c7b4d48e37d81245851477a3e/packages/cactus-plugin-satp-hermes/src/test/typescript/unit/) test files (client and server side along with the recovery procedure).
 
 ## Usage

diff --git a/packages/cactus-plugin-satp-hermes/package.json b/packages/cactus-plugin-satp-hermes/package.json
@@ -40,6 +40,10 @@
     {
       "name": "Bruno Mateus",
       "url": "https://github.com/brunoffmateus"
+    },
+    {
+      "name": "Yogesh D",
+      "url": "https://github.com/Yogesh01000100"
     }
   ],
   "main": "dist/lib/main/typescript/index.js",
@@ -79,6 +83,8 @@
     "preinstall": "curl -L https://foundry.paradigm.xyz | bash && foundryup",
     "pretsc": "npm run generate-sdk",
     "start-gateway": "node ./dist/lib/main/typescript/plugin-satp-hermes-gateway-cli.js",
+    "test:integration": "npx jest ./src/test/typescript/integration",
+    "test:unit": "npx jest ./src/test/typescript/unit",
     "tsc": "tsc --project ./tsconfig.json",
     "watch": "tsc --build --watch",
     "db:setup": "bash -c 'npm run db:destroy || true && run-s db:start db:migrate db:seed'",
@@ -136,6 +142,7 @@
     "jsonc": "2.0.0",
     "knex": "2.4.0",
     "kubo-rpc-client": "3.0.1",
+    "node-schedule": "2.1.1",
     "npm-run-all": "4.1.5",
     "openzeppelin-solidity": "3.4.2",
     "pg": "8.13.1",
@@ -163,6 +170,7 @@
     "@types/fs-extra": "11.0.4",
     "@types/google-protobuf": "3.15.12",
     "@types/node": "18.18.2",
+    "@types/node-schedule": "2.1.7",
     "@types/pg": "8.11.10",
     "@types/swagger-ui-express": "4.1.6",
     "@types/tape": "4.13.4",

diff --git a/packages/cactus-plugin-satp-hermes/src/knex/knexfile-remote.ts b/packages/cactus-plugin-satp-hermes/src/knex/knexfile-remote.ts
@@ -6,11 +6,11 @@ import { Knex } from "knex";
 const envPath = process.env.ENV_PATH;
 dotenv.config({ path: envPath });
 
-const config: { [key: string]: Knex.Config } = {
-  development: {
+export const knexRemoteInstance: { [key: string]: Knex.Config } = {
+  default: {
     client: "sqlite3",
     connection: {
-      filename: path.resolve(__dirname, ".dev.remote-" + uuidv4() + ".sqlite3"),
+      filename: path.resolve(__dirname, `.dev.remote-${uuidv4()}.sqlite3`),
     },
     migrations: {
       directory: path.resolve(__dirname, "migrations"),
@@ -31,5 +31,3 @@ const config: { [key: string]: Knex.Config } = {
     },
   },
 };
-
-export default config;
diff --git a/packages/cactus-plugin-satp-hermes/src/knex/knexfile.ts b/packages/cactus-plugin-satp-hermes/src/knex/knexfile.ts
@@ -6,8 +6,8 @@ import { Knex } from "knex";
 const envPath = process.env.ENV_PATH;
 dotenv.config({ path: envPath });
 
-const config: { [key: string]: Knex.Config } = {
-  development: {
+export const knexLocalInstance: { [key: string]: Knex.Config } = {
+  default: {
     client: "sqlite3",
     connection: {
       filename: path.resolve(__dirname, `.dev.local-${uuidv4()}.sqlite3`),
@@ -34,5 +34,3 @@ const config: { [key: string]: Knex.Config } = {
     },
   },
 };
-
-export default config;
diff --git a/packages/cactus-plugin-satp-hermes/src/knex/migrations/20220331132128_create_logs_table.ts b/packages/cactus-plugin-satp-hermes/src/knex/migrations/20220331132128_create_logs_table.ts
@@ -2,7 +2,7 @@ import { Knex } from "knex";
 
 export function up(knex: Knex): Knex.SchemaBuilder {
   return knex.schema.createTable("logs", (table) => {
-    table.string("sessionID").notNullable();
+    table.string("sessionId").notNullable();
     table.string("type").notNullable();
     table.string("key").notNullable().primary();
     table.string("operation").notNullable();

diff --git a/packages/cactus-plugin-satp-hermes/src/main/proto/cacti/satp/v02/common/session.proto b/packages/cactus-plugin-satp-hermes/src/main/proto/cacti/satp/v02/common/session.proto
@@ -80,6 +80,7 @@ message SessionData {
     cacti.satp.v02.common.MessageType phase_error = 68;
     bool recovered_tried = 69;
     SATPMessages satp_Messages = 70; 
+    Type role = 71;
 }
 
 enum State {
@@ -93,6 +94,12 @@ enum State {
     STATE_RECOVERING = 7;
 }
 
+enum Type {
+    UNKNOWN = 0;
+    CLIENT = 1;
+    SERVER = 2;
+}
+
 message SATPMessages {
     Stage0Messages stage0 = 1;
     Stage1Messages stage1 = 2;

diff --git a/packages/cactus-plugin-satp-hermes/src/main/proto/cacti/satp/v02/crash_recovery.proto b/packages/cactus-plugin-satp-hermes/src/main/proto/cacti/satp/v02/crash_recovery.proto
@@ -9,7 +9,7 @@ service CrashRecovery   {
 
   // step RPCs
   rpc RecoverV2Message(RecoverMessage) returns (RecoverUpdateMessage);
-  rpc RecoverV2SuccessMessage(RecoverSuccessMessage) returns (google.protobuf.Empty);
+  rpc RecoverV2SuccessMessage(RecoverSuccessMessage) returns (RecoverSuccessMessageResponse);
   rpc RollbackV2Message(RollbackMessage) returns (RollbackAckMessage);
 }
 
@@ -21,15 +21,15 @@ message RecoverMessage {
   bool is_backup = 5;
   string new_identity_public_key = 6;
   int64 last_entry_timestamp = 7;
-  string sender_signature = 8;
+  string client_signature = 8;
 }
 
 message RecoverUpdateMessage {
   string session_id = 1;
   string message_type = 2;
   string hash_recover_message = 3;
-  repeated LocalLog recovered_logs = 4;
-  string sender_signature = 5;
+  repeated persistLogEntry recovered_logs = 4;
+  string server_signature = 5;
 }
 
 message RecoverSuccessMessage {
@@ -38,7 +38,13 @@ message RecoverSuccessMessage {
   string hash_recover_update_message = 3;
   bool success = 4;
   repeated string entries_changed = 5;
-  string sender_signature = 6;
+  string client_signature = 6;
+}
+
+message RecoverSuccessMessageResponse {
+  string session_id = 1;
+  bool received = 2;
+  string server_signature = 3;
 }
 
 message RollbackMessage {
@@ -47,7 +53,7 @@ message RollbackMessage {
   bool success = 3;
   repeated string actions_performed = 4;
   repeated string proofs = 5;
-  string sender_signature = 6;
+  string client_signature = 6;
 }
 
 message RollbackAckMessage {
@@ -56,10 +62,10 @@ message RollbackAckMessage {
   bool success = 3;
   repeated string actions_performed = 4;
   repeated string proofs = 5;
-  string sender_signature = 6;
+  string server_signature = 6;
 }
 
-message LocalLog {
+message persistLogEntry {
   string session_id = 1;
   string type = 2;
   string key = 3;
@@ -74,16 +80,14 @@ message RollbackLogEntry {
   string stage = 2;
   string timestamp = 3;
   string action = 4; // action performed during rollback
-  string status = 5; // status of rollback (e.g., SUCCESS, FAILED)
-  string details = 6; // Additional details or metadata about the rollback
+  string status = 5; 
+  string details = 6;
 }
 
 message RollbackState {
   string session_id = 1;
   string current_stage = 2;
-  int32 steps_remaining = 3;
-  repeated RollbackLogEntry rollback_log_entries = 4;
-  string estimated_time_to_completion = 5;
-  string status = 6; // Overall status (e.g., IN_PROGRESS, COMPLETED, FAILED)
-  string details = 7; // Additional metadata or information
+  repeated RollbackLogEntry rollback_log_entries = 3;
+  string status = 4; // Overall status (e.g., IN_PROGRESS, COMPLETED, FAILED)
+  string details = 5;
 }
diff --git a/...actus-plugin-satp-hermes/src/main/typescript/blo/admin/get-healthcheck-handler-service.ts b/...actus-plugin-satp-hermes/src/main/typescript/blo/admin/get-healthcheck-handler-service.ts
@@ -41,7 +41,7 @@ export async function getHealthCheckService(
   const status = manager.healthCheck();
 
   const res: HealthCheckResponse = {
-    status: status
+    status: status,
   };
 
   log.debug(res);

diff --git a/...cactus-plugin-satp-hermes/src/main/typescript/blo/transaction/transact-handler-service.ts b/...cactus-plugin-satp-hermes/src/main/typescript/blo/transaction/transact-handler-service.ts
@@ -1,4 +1,3 @@
-import { Logger } from "@hyperledger/cactus-common";
 import { TransactRequest, TransactResponse } from "../../public-api";
 import { SATPManager } from "../../gol/satp-manager";
 import { populateClientSessionData } from "../../core/session-utils";
@@ -12,7 +11,6 @@ import { GatewayOrchestrator } from "../../gol/gateway-orchestrator";
 import { GatewayIdentity } from "../../core/types";
 import { SATP_VERSION } from "../../core/constants";
 import { getStatusService } from "../admin/get-status-handler-service";
-import { log } from "console";
 
 // todo
 export async function executeTransact(
@@ -26,7 +24,7 @@ export async function executeTransact(
     label: fnTag,
     level: logLevel,
   });
-  
+
   logger.info(`${fnTag}, executing transaction endpoint`);
 
   //TODO check input for valid strings...