Merge pull request #1093 from skalenetwork/2.7.0-stable

2.7.0 stable

.github/workflows/test.yml

@@ -8,47 +8,74 @@ env:
   SGX_WALLET_TAG: "1.83.0-beta.5"
   CODECOV_TOKEN: ${{ secrets.CODECOV_TOKEN }}
   PYTHON_VERSION: 3.11
+
 jobs:
   test_core:
     runs-on: ubuntu-latest
     env:
       ETH_PRIVATE_KEY: ${{ secrets.ETH_PRIVATE_KEY }}
       ENDPOINT: http://127.0.0.1:8545
       CODECOV_TOKEN: ${{ secrets.CODECOV_TOKEN }}
-      IMA_ENDPOINT: ${{ secrets.IMA_ENDPOINT }}
       SCHAIN_TYPE: ${{ secrets.SCHAIN_TYPE }}
     steps:
       - uses: actions/checkout@v2
         with:
           submodules: true
+
       - name: Set up Python ${{ env.PYTHON_VERSION }}
         uses: actions/setup-python@v1
         with:
           python-version: ${{ env.PYTHON_VERSION }}
+
       - name: Install python dependencies
         run: bash ./scripts/install_python_dependencies.sh
+
       - name: Lint with flake8
         run: flake8 .
+
       - name: Launch anvil node
         run: |
           docker run -d --network host --name anvil ghcr.io/foundry-rs/foundry anvil && sleep 5 && docker logs anvil --tail 1000
+
       - name: Deploy manager & ima contracts
         run: |
           bash ./helper-scripts/deploy_test_ima.sh
+
+      - name: Cleanup skale-manager image
+        run: |
           docker rmi -f skalenetwork/skale-manager:${{ env.MANAGER_TAG }}
+
       - name: Show stats before tests
         if: always()
         run: | 
           sudo lsblk -f
           sudo free -h
+
       - name: Run core tests
         run: |
           bash ./scripts/run_core_tests.sh
-      - name: Show stats after tests
+
+      - name: Cleanup docker artifacts
+        run: |
+          docker rm -f $(docker ps -aq)
+          docker rmi -f $(docker images -q)
+
+      - name: Show stats after core tests
+        if: always()
+        run: |
+          sudo lsblk -f
+          sudo free -h
+
+      - name: Run firewall tests
+        run: |
+          bash ./scripts/run_firewall_test.sh
+
+      - name: Show stats after firewall tests
         if: always()
         run: | 
           sudo lsblk -f
           sudo free -h
+
       - name: Run codecov
         run: |
           codecov -t $CODECOV_TOKEN

.gitignore

@@ -63,6 +63,8 @@ schain_status_sample.json
 *.lock
 
 skale-manager-*
+manager.json
+meta.json
 
 tests/skale-data/contracts_info/ima.json
 tests/skale-data/contracts_info/schain_ima_abi.json

VERSION

@@ -1 +1 @@
-2.6.1
+2.7.0

admin.py

@@ -27,7 +27,7 @@
 from core.schains.process_manager import run_process_manager
 from core.schains.cleaner import run_cleaner
 from core.updates import soft_updates
-from core.filebeat import update_filebeat_service
+from core.monitoring import update_monitoring_services
 
 from tools.configs import BACKUP_RUN, INIT_LOCK_PATH, PULL_CONFIG_FOR_SCHAIN
 from tools.configs.web3 import (
@@ -82,7 +82,7 @@ def worker():
     skale_ima = SkaleIma(ENDPOINT, MAINNET_IMA_ABI_FILEPATH, wallet)
     if BACKUP_RUN:
         logger.info('Running sChains in snapshot download mode')
-    update_filebeat_service(node_config.ip, node_config.id, skale)
+    update_monitoring_services(node_config.ip, node_config.id, skale)
     monitor(skale, skale_ima, node_config)
 
 

app.py

@@ -35,10 +35,10 @@
     FLASK_DEBUG_MODE
 )
 from tools.configs.web3 import ENDPOINT
-from tools.db import get_database, REDIS_URI
 from tools.docker_utils import DockerUtils
 from tools.helper import wait_until_admin_inited
 from tools.logger import init_api_logger
+from tools.resources import get_database, REDIS_URI
 from tools.str_formatters import arguments_list_string
 
 from web.routes.node import node_bp

core/monitoring.py

@@ -0,0 +1,121 @@
+#   -*- coding: utf-8 -*-
+#
+#   This file is part of SKALE Admin
+#
+#   Copyright (C) 2020 SKALE Labs
+#
+#   This program is free software: you can redistribute it and/or modify
+#   it under the terms of the GNU Affero General Public License as published by
+#   the Free Software Foundation, either version 3 of the License, or
+#   (at your option) any later version.
+#
+#   This program is distributed in the hope that it will be useful,
+#   but WITHOUT ANY WARRANTY; without even the implied warranty of
+#   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+#   GNU Affero General Public License for more details.
+#
+#   You should have received a copy of the GNU Affero General Public License
+#   along with this program.  If not, see <https://www.gnu.org/licenses/>.
+
+import logging
+from typing import Optional
+
+from tools.helper import process_template
+from tools.docker_utils import DockerUtils
+
+from tools.configs import SKALE_DIR_HOST
+from tools.configs.monitoring import (
+    FILEBEAT_TEMPLATE_PATH, FILEBEAT_CONTAINER_NAME,
+    FILEBEAT_CONFIG_PATH,
+    INFLUX_URL,
+    TELEGRAF,
+    TELEGRAF_CONTAINER_NAME, TELEGRAF_IMAGE,
+    TELEGRAF_TEMPLATE_PATH,
+    TELEGRAF_CONFIG_PATH,
+    TELEGRAF_MEM_LIMIT
+)
+
+logger = logging.getLogger(__name__)
+
+
+class TelegrafNotConfiguredError(Exception):
+    pass
+
+
+def update_filebeat_service(node_ip, node_id, skale, dutils: Optional[DockerUtils] = None):
+    dutils = dutils or DockerUtils()
+    contract_address = skale.manager.address
+    template_data = {
+        'ip': node_ip,
+        'id': node_id,
+        'contract_address': contract_address
+    }
+
+    logger.info('Configuring filebeat %s', template_data)
+    process_template(FILEBEAT_TEMPLATE_PATH, FILEBEAT_CONFIG_PATH, template_data)
+    filebeat_container = dutils.client.containers.get(FILEBEAT_CONTAINER_NAME)
+    filebeat_container.restart()
+    logger.info('Filebeat config updated, telegraf restarted')
+
+
+def filebeat_config_processed() -> bool:
+    with open(FILEBEAT_CONFIG_PATH) as f:
+        return 'id: ' in f.read()
+
+
+def ensure_telegraf_running(dutils: Optional[DockerUtils] = None) -> None:
+    dutils = dutils or DockerUtils()
+    if dutils.is_container_exists(TELEGRAF_CONTAINER_NAME):
+        dutils.restart(TELEGRAF_CONTAINER_NAME)
+    else:
+        dutils.run_container(
+            image_name=TELEGRAF_IMAGE,
+            name=TELEGRAF_CONTAINER_NAME,
+            network_mode='host',
+            user='telegraf:998',
+            restart_policy={'name': 'on-failure'},
+            environment={'HOST_PROC': '/host/proc'},
+            volumes={
+                '/proc': {'bind': '/host/proc', 'mode': 'ro'},
+                f'{SKALE_DIR_HOST}/config/telegraf.conf': {'bind': '/etc/telegraf/telegraf.conf', 'mode': 'ro'},  # noqa
+                f'{SKALE_DIR_HOST}/node_data/telegraf': {'bind': '/var/lib/telegraf', 'mode': 'rw'},
+                '/var/run/skale/': {'bind': '/var/run/skale', 'mode': 'rw'}
+            },
+            mem_limit=TELEGRAF_MEM_LIMIT
+        )
+
+
+def update_telegraf_service(
+    node_ip: str,
+    node_id: int,
+    url: str = INFLUX_URL,
+    dutils: Optional[DockerUtils] = None
+) -> None:
+    dutils = dutils or DockerUtils()
+    template_data = {
+        'ip': node_ip,
+        'node_id': str(node_id),
+        'url': url
+    }
+    missing = list(filter(lambda k: not template_data[k], template_data))
+
+    if missing:
+        emsg = f'TELEGRAF=True is set, but missing options {template_data}'
+        raise TelegrafNotConfiguredError(emsg)
+
+    logger.info('Configuring telegraf %s', template_data)
+    process_template(TELEGRAF_TEMPLATE_PATH, TELEGRAF_CONFIG_PATH, template_data)
+
+    ensure_telegraf_running(dutils)
+    logger.info('Telegraf config updated, telegraf restarted')
+
+
+def telegraf_config_processed() -> bool:
+    with open(TELEGRAF_CONFIG_PATH) as f:
+        return 'id: ' in f.read()
+
+
+def update_monitoring_services(node_ip, node_id, skale, dutils: Optional[DockerUtils] = None):
+    update_filebeat_service(node_ip, node_id, skale, dutils=dutils)
+    if TELEGRAF:
+        update_telegraf_service(node_ip, node_id, dutils=dutils)

core/node.py

@@ -37,7 +37,7 @@
 from skale.utils.helper import ip_from_bytes
 from skale.utils.web3_utils import public_key_to_address, to_checksum_address
 
-from core.filebeat import update_filebeat_service
+from core.monitoring import update_monitoring_services
 from tools.configs import WATCHDOG_PORT, CHANGE_IP_DELAY, CHECK_REPORT_PATH, META_FILEPATH
 from tools.helper import read_json
 from tools.str_formatters import arguments_list_string
@@ -135,7 +135,7 @@ def register(self, ip, public_ip, port, name, domain_name,
         self.config.name = name
         self.config.ip = ip
 
-        update_filebeat_service(public_ip, self.config.id, self.skale)
+        update_monitoring_services(public_ip, self.config.id, self.skale)
         return self._ok(data=self.config.all())
 
     def create_node_on_contracts(self, ip, public_ip, port, name, domain_name,