add checker

checker.py

@@ -0,0 +1,152 @@
+import re
+from datetime import datetime, timedelta
+import subprocess
+import os
+import sys
+import pkg_resources
+
+required_packages = ['grist-api', 'colorama']
+installed_packages = [pkg.key for pkg in pkg_resources.working_set]
+
+for package in required_packages:
+    if package not in installed_packages:
+        subprocess.check_call([sys.executable, '-m', 'pip', 'install', package, '--break-system-packages'])
+
+from grist_api import GristDocAPI
+import colorama
+import requests
+import logging
+import socket
+
+
+class GRIST:
+    def __init__(self, server, doc_id, api_key, logger):
+        self.server = server
+        self.doc_id = doc_id
+        self.api_key = api_key
+        self.logger = logger
+        self.grist = GristDocAPI(doc_id, server=server, api_key=api_key)
+
+    def table_name_convert(self, table_name):
+        return table_name.replace(" ", "_")
+
+    def to_timestamp(self, dtime: datetime) -> int:
+        if dtime.tzinfo is None:
+            dtime = dtime.replace(tzinfo=timezone(timedelta(hours=3))) 
+        return int(dtime.timestamp())
+
+    def insert_row(self, data, table):
+        data = {key.replace(" ", "_"): value for key, value in data.items()}
+        row_id = self.grist.add_records(self.table_name_convert(table), [data])
+        return row_id
+
+    def update_column(self, row_id, column_name, value, table):
+        if isinstance(value, datetime):
+            value = self.to_timestamp(value)
+        column_name = column_name.replace(" ", "_")
+        self.grist.update_records(self.table_name_convert(table), [{ "id": row_id, column_name: value }])
+
+    def delete_row(self, row_id, table):
+        self.grist.delete_records(self.table_name_convert(table), [row_id])
+
+    def update(self, row_id, updates, table):
+        for column_name, value in updates.items():
+            if isinstance(value, datetime):
+                updates[column_name] = self.to_timestamp(value)
+        updates = {column_name.replace(" ", "_"): value for column_name, value in updates.items()}
+        self.grist.update_records(self.table_name_convert(table), [{"id": row_id, **updates}])
+
+    def fetch_table(self, table):
+        return self.grist.fetch_table(self.table_name_convert(table))
+
+    def find_record(self, id=None, state=None, name=None, table=None):
+        if table is None:
+            raise ValueError("Table is not specified")
+        table_data = self.grist.fetch_table(self.table_name_convert(table))
+        if id is not None:
+            record = [row for row in table_data if row.id == id]
+            return record
+        if state is not None and name is not None:
+            record = [row for row in table_data if row.State == state and row.name == name]
+            return record
+        if state is not None:
+            record = [row for row in table_data if row.State == state]
+            return record
+        if name is not None:
+            record = [row for row in table_data if row.Name == name]
+            return record
+
+    def find_settings(self, key, table):
+        table = self.fetch_table(self.table_name_convert(table))
+        for record in table:
+            if record.Setting == key:
+                if record.Value is None or record.Value == "":
+                    raise ValueError(f"Setting {key} blank")
+                return record.Value
+        raise ValueError(f"Setting {key} not found")
+
+
+def check_logs():
+    # Initialize counters
+    error_count = 0
+    sync_count = 0
+    total_challenges = 0
+    
+    # Get current time and 24 hours ago
+    current_time = datetime.now()
+    day_ago = current_time - timedelta(days=1)
+    
+    try:
+        result = subprocess.run(['docker', 'compose', 'logs'], cwd='/root/node/', capture_output=True, text=True)
+        log_content = result.stdout
+    except subprocess.CalledProcessError as e:
+        raise Exception(f"Error running docker compose logs: {e}")
+    
+    for line in log_content.split('\n'):
+        timestamp_match = re.search(r'(\d{4}-\d{2}-\d{2}T\d{2}:\d{2}:\d{2})', line)
+        if timestamp_match: timestamp = datetime.strptime(timestamp_match.group(1), '%Y-%m-%dT%H:%M:%S')
+        else: timestamp = None
+        
+        if not timestamp: continue
+        if timestamp < day_ago: continue
+        if "Error from tendermint rpc" in line: error_count += 1
+        if "Synced with network" in line: sync_count += 1
+            
+        challenge_match = re.search(r'made (\d+) secret challenges', line)
+        if challenge_match: total_challenges += int(challenge_match.group(1))
+    
+    return {
+        "rpc_errors": error_count,
+        "sync_events": sync_count, 
+        "total_challenges": total_challenges
+    }
+
+if __name__ == "__main__":
+    colorama.init(autoreset=True)
+    logger = logging.getLogger("Impact updater")
+    logger.setLevel(logging.INFO)
+    formatter = logging.Formatter("%(asctime)s - %(name)s - %(levelname)s - %(message)s")
+    ch = logging.StreamHandler()
+    ch.setFormatter(formatter)
+    logger.addHandler(ch)
+
+    grist_server = "###GRIST_SERVER###"
+    grist_doc_id = "###GRIST_DOC_ID###"
+    grist_api_key = "###GRIST_API_KEY###"
+    grist_row_name = socket.gethostname()
+    nodes_table = "Nodes"
+    grist = GRIST(grist_server, grist_doc_id, grist_api_key, logger)
+    vms = grist.find_record(name=grist_row_name, table=nodes_table)
+    current_vm = vms[0]
+    def grist_callback(msg): grist.update(current_vm.id, msg, nodes_table)
+    #register_url = grist.find_settings("Nillion checker url", "Settings")
+
+    try:        
+        result = check_logs()
+        data = f"{result['sync_events']}/{result['total_challenges']}/{result['rpc_errors']}" # Syncs/Challenges/RPC errors
+        grist_callback({ "Health": data })
+        print(result)
+    except Exception as e:  
+        logger.error(f"Error: {e}")
+        grist_callback({ "Health": f"Error: {e}" })
+    

playbook.yml

@@ -174,6 +174,9 @@
         ./update.sh PRIVATE "{{ private_key }}"
         ./update.sh PUBLIC "{{ public_key }}"
         ./update.sh RPC "{{ rpc_url }}"
+        ./update.sh GRIST_SERVER "{{ grist_server }}"
+        ./update.sh GRIST_DOC_ID "{{ grist_doc_id }}"
+        ./update.sh GRIST_API_KEY "{{ grist_api_key }}"
       args:
         chdir: "{{ ansible_env.HOME }}/node"
       changed_when: false
@@ -316,6 +319,36 @@
       async: "{{ 60 * 80 }}"
       poll: "{{ 60 }}"
 
+    - name: Copy checker service file
+      ansible.builtin.copy:
+        dest: /etc/systemd/system/nillion-checker.service
+        content: |
+          [Unit]
+          Description=Nillion Checker Service
+          After=network.target
+
+          [Service]
+          Type=simple
+          User=root
+          WorkingDirectory={{ ansible_env.HOME }}/node
+          ExecStart=/usr/bin/python3 {{ ansible_env.HOME }}/node/checker.py
+          Restart=always
+          RestartSec=1800
+
+          [Install]
+          WantedBy=multi-user.target
+        mode: '0644'
+
+    - name: Reload systemd
+      ansible.builtin.systemd:
+        daemon_reload: yes
+
+    - name: Enable and start nillion-checker service
+      ansible.builtin.systemd:
+        name: nillion-checker
+        enabled: yes
+        state: started
+
 
 #    - name: Check "not have enough balance"
 #      ansible.builtin.command: docker logs {{ item }} 2>&1

update.sh

@@ -12,6 +12,7 @@ NEW_VALUE=$2
 FILES=(
     "credentials.json"
     "docker-compose.yml"
+    "checker.py"
 )
 
 for FILE in "${FILES[@]}"; do