nillion/checker.py

# flake8: noqa
# pylint: disable=broad-exception-raised, raise-missing-from, too-many-arguments, redefined-outer-name
# pylance: disable=reportMissingImports, reportMissingModuleSource, reportGeneralTypeIssues
# type: ignore

import re
from datetime import datetime, timedelta, timezone
import subprocess
import os
import time
import random
import sys
import pkg_resources
import requests

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 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(log_handler):
    # Initialize counters
    error_count = 0
    sync_count = 0
    total_challenges = 0

    try:
        logs = subprocess.run(['docker', 'compose', 'logs', '--since', '24h'], cwd='/root/node/', capture_output=True, text=True, check=True)
        log_content = logs.stdout
    except subprocess.CalledProcessError as e:
        raise RuntimeError(f"Error running docker compose logs: {e}") from e
    
    for line in log_content.split('\n'):
        if "Error from tendermint rpc" in line: 
            error_count += 1
            log_handler.error(f"RPC error: {line}")
        if "Is your verifier's account funded" in line:
            log_handler.error(f"Verifier account not funded: {line}")
            error_count += 1
        if "Synced with network" in line:
            sync_count += 1
            log_handler.info(f"Synced with network: {line}")

        challenge_match = re.search(r'made (\d+) secret challenges', line)
        if challenge_match:
            total_challenges += int(challenge_match.group(1))
            log_handler.info(f"Made {total_challenges} secret challenges: {line}")
    data = {
        "errors": error_count,
        "sync_events": sync_count, 
        "total_challenges": total_challenges
    }
    log_handler.info(f"Result: {data}")
    return data

if __name__ == "__main__":
    print("Checker started")
    colorama.init(autoreset=True)
    logger = logging.getLogger("Checker")
    logger.setLevel(logging.INFO)
    formatter = logging.Formatter("%(asctime)s - %(name)s - %(levelname)s - %(message)s")
    ch = logging.StreamHandler()
    ch.setFormatter(formatter)
    logger.addHandler(ch)

    random_sleep = random.randint(1, 600)
    logger.info(f"Sleeping for {random_sleep} seconds")
    time.sleep(random_sleep)

    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)
    current_vm = grist.find_record(name=GRIST_ROW_NAME, table=NODES_TABLE)[0]
    def grist_callback(msg): grist.update(current_vm.id, msg, NODES_TABLE)

    for attempt in range(3):
        try:        
            result = check_logs(logger)
            data = f"{result['sync_events']}/{result['total_challenges']}/{result['errors']}" # Syncs/Challenges/Errors
            grist_callback({ "Health": data })
            print(result)
            break
        except Exception as e:  
            logger.error(f"Error on attempt {attempt+1}/3: {e}")
            if attempt == 2:
                grist_callback({ "Health": f"Error: {e}" })
            if attempt < 2:
                time.sleep(5)