diff --git a/README.md b/README.md
index dc11297..8798f86 100644
--- a/README.md
+++ b/README.md
@@ -13,28 +13,37 @@ python ingest_XXXX.py [options]
```
**Note:** The `` can be a file or a directory of files, depending on the ingestion script.
-###### Options
-| Argument | Description |
-| --------------- | -------------------------------------------------------------------------------------------- |
-| `--dry-run` | Perform a dry run without indexing records to Elasticsearch. |
-| `--batch_size` | Number of records to index in a batch *(default 25,000)*. |
+###
+###### General arguments
+| Argument | Description |
+|-------------------|----------------------------------------------------------------|
+| `input_path` | Path to the input file or directory |
+| `--dry-run` | Dry run *(do not index records to Elasticsearch)* |
+| `--watch` | Watch the input file for new lines and index them in real time |
-###### Elasticsearch Connnection Options
-| Argument | Description |
-| --------------- | -------------------------------------------------------------------------------------------- |
-| `--host` | Elasticsearch host *(default 'localhost')*. |
-| `--port` | Elasticsearch port *(default 9200)*. |
-| `--user` | Elasticsearch username *(default 'elastic')*. |
-| `--password` | Elasticsearch password. If not provided, it checks the environment variable **ES_PASSWORD**. |
-| `--api-key` | Elasticsearch API Key for authentication. |
-| `--self-signed` | Allow self-signed certificates. |
+###### Elasticsearch arguments
+| Argument | Description | Default |
+|-------------------|--------------------------------------------------------------------------------------|---------------|
+| `--host` | Elasticsearch host *(Will sniff for other nodes in the cluster)* | `localhost` |
+| `--port` | Elasticsearch port | `9200` |
+| `--user` | Elasticsearch username | `elastic` |
+| `--password` | Elasticsearch password *(if not provided, check environment variable `ES_PASSWORD`)* | |
+| `--api-key` | Elasticsearch API Key for authentication | |
+| `--self-signed` | Elastic search instance is using a self-signed certificate | `true` |
+| `--index` | Elasticsearch index name | `masscan-logs`|
+| `--shards` | Number of shards for the index | `1` |
+| `--replicas` | Number of replicas for the index | `1` |
-###### Elasticsearch Index Options
-| Argument | Description |
-| --------------- | -------------------------------------------------------------------------------------------- |
-| `--index` | Elasticsearch index name *(default 'zone_files')*. |
-| `--replicas` | Number of replicas for the index. |
-| `--shards` | Number of shards for the index |
+###### Performance arguments
+| Argument | Description | Default |
+|-------------------|--------------------------------------------------------------------------------------|---------------|
+| `--batch-max` | Maximum size in MB of a batch | `10` |
+| `--batch-size` | Number of records to index in a batch | `5000` |
+| `--batch-threads` | Number of threads to use when indexing in batches | `2` |
+| `--retries` | Number of times to retry indexing a batch before failing | `10` |
+| `--timeout` | Number of seconds to wait before retrying a batch | `30` |
+
+**NOTE:** Using `--batch-threads` as 4 and `--batch-size` as 10000 with 3 nodes would process 120,000 records before indexing 40,000 per node.
___
diff --git a/ingestors/__pycache__/sniff_patch.cpython-312.pyc b/ingestors/__pycache__/sniff_patch.cpython-312.pyc
new file mode 100644
index 0000000..d3fc380
Binary files /dev/null and b/ingestors/__pycache__/sniff_patch.cpython-312.pyc differ
diff --git a/ingestors/ingest_httpx.py b/ingestors/ingest_httpx.py
index 9ded692..91a851e 100644
--- a/ingestors/ingest_httpx.py
+++ b/ingestors/ingest_httpx.py
@@ -12,9 +12,10 @@ import argparse
import json
import logging
import os
+import time
try:
- from elasticsearch import Elasticsearch, helpers
+ from elasticsearch import Elasticsearch, helpers, ConnectionError, TransportError
except ImportError:
raise ImportError('Missing required \'elasticsearch\' library. (pip install elasticsearch)')
@@ -105,6 +106,7 @@ class ElasticIndexer:
}
'''
+ count = 0
records = []
with open(file_path, 'r') as file:
@@ -126,15 +128,29 @@ class ElasticIndexer:
else:
record = {'_index': self.es_index, '_source': record}
records.append(record)
-
+ count += 1
if len(records) >= batch_size:
- success, _ = helpers.bulk(self.es, records)
- logging.info(f'Successfully indexed {success} records to {self.es_index}')
- records = []
+ while True:
+ try:
+ success, _ = helpers.bulk(self.es, records)
+ except (ConnectionError, TransportError) as e:
+ logging.error(f'Failed to index records to Elasticsearch. ({e})')
+ time.sleep(60)
+ else:
+ logging.info(f'Successfully indexed {success:,} ({count:,}) records to {self.es_index} from {file_path}')
+ records = []
+ break
if records:
- success, _ = helpers.bulk(self.es, records)
- logging.info(f'Successfully indexed {success} records to {self.es_index}')
+ while True:
+ try:
+ success, _ = helpers.bulk(self.es, records)
+ except (ConnectionError, TransportError) as e:
+ logging.error(f'Failed to index records to Elasticsearch. ({e})')
+ time.sleep(60)
+ else:
+ logging.info(f'Successfully indexed {success:,} ({count:,}) records to {self.es_index} from {file_path}')
+ break
def main():
diff --git a/ingestors/ingest_masscan.py b/ingestors/ingest_masscan.py
index c1dbf15..79df82a 100644
--- a/ingestors/ingest_masscan.py
+++ b/ingestors/ingest_masscan.py
@@ -5,8 +5,14 @@
#
# This script takes JSON formatted masscan logs with banners and indexes them into Elasticsearch.
#
-# Saving my "typical" masscan command here for reference to myself:
-# masscan 0.0.0.0/0 -p3559,1900 --banners --open-only --rate 25000 --excludefile exclude.conf -oJ output.json --interactive
+# Saving my "typical" masscan setup & command here for reference to myself:
+# apt-get install iptables masscan libpcap-dev screen
+# /sbin/iptables -A INPUT -p tcp --dport 61010 -j DROP
+# printf "0.0.0.0/8\n10.0.0.0/8\n100.64.0.0/10\n127.0.0.0/8\n169.254.0.0/16\n172.16.0.0/12\n192.0.0.0/24\n192.0.0.0/29\n192.0.0.170/32\n192.0.0.171/32\n192.0.2.0/24\n192.88.99.0/24\n192.168.0.0/16\n198.18.0.0/15\n198.51.100.0/24\n203.0.113.0/24\n240.0.0.0/4\n255.255.255.255/32\n" > exclude.conf
+# screen -S scan
+# masscan 0.0.0.0/0 -p8080,8888,8000 --banners --source-port 61010 --open-only --rate 35000 --excludefile exclude.conf -oJ output_new.json --interactive
+#
+# Note: The above iptables rule is not persistent and will be removed on reboot.
import argparse
import json
@@ -26,11 +32,11 @@ logging.basicConfig(level=logging.INFO, format='%(asctime)s - %(name)s - %(level
class ElasticIndexer:
- def __init__(self, es_host: str, es_port: str, es_user: str, es_password: str, es_api_key: str, es_index: str, dry_run: bool = False, self_signed: bool = False):
+ def __init__(self, es_host: str, es_port: int, es_user: str, es_password: str, es_api_key: str, es_index: str, dry_run: bool = False, self_signed: bool = False, retries: int = 10, timeout: int = 30):
'''
Initialize the Elastic Search indexer.
- :param es_host: Elasticsearch host
+ :param es_host: Elasticsearch host(s)
:param es_port: Elasticsearch port
:param es_user: Elasticsearch username
:param es_password: Elasticsearch password
@@ -38,6 +44,8 @@ class ElasticIndexer:
:param es_index: Elasticsearch index name
:param dry_run: If True, do not initialize Elasticsearch client
:param self_signed: If True, do not verify SSL certificates
+ :param retries: Number of times to retry indexing a batch before failing
+ :param timeout: Number of seconds to wait before retrying a batch
'''
self.dry_run = dry_run
@@ -45,10 +53,24 @@ class ElasticIndexer:
self.es_index = es_index
if not dry_run:
+ es_config = {
+ 'hosts': [f'{es_host}:{es_port}'],
+ 'verify_certs': self_signed,
+ 'ssl_show_warn': self_signed,
+ 'request_timeout': timeout,
+ 'max_retries': retries,
+ 'retry_on_timeout': True,
+ 'sniff_on_start': True,
+ 'sniff_on_node_failure': True,
+ 'min_delay_between_sniffing': 60
+ }
+
if es_api_key:
- self.es = Elasticsearch([f'{es_host}:{es_port}'], headers={'Authorization': f'ApiKey {es_api_key}'}, verify_certs=self_signed, ssl_show_warn=self_signed)
+ es_config['headers'] = {'Authorization': f'ApiKey {es_api_key}'}
else:
- self.es = Elasticsearch([f'{es_host}:{es_port}'], basic_auth=(es_user, es_password), verify_certs=self_signed, ssl_show_warn=self_signed)
+ es_config['basic_auth'] = (es_user, es_password)
+
+ self.es = Elasticsearch(**es_config)
def create_index(self, shards: int = 1, replicas: int = 1):
@@ -84,13 +106,23 @@ class ElasticIndexer:
logging.warning(f'Index \'{self.es_index}\' already exists.')
- def process_file(self, file_path: str, batch_size: int):
+ def get_cluster_size(self) -> int:
+ '''Get the number of nodes in the Elasticsearch cluster.'''
+
+ cluster_stats = self.es.cluster.stats()
+ number_of_nodes = cluster_stats['nodes']['count']['total']
+
+ return number_of_nodes
+
+
+ def process_file(self, file_path: str, watch: bool = False, chunk: dict = {}):
'''
Read and index Masscan records in batches to Elasticsearch, handling large volumes efficiently.
:param file_path: Path to the Masscan log file
:param batch_size: Number of records to process before indexing
-
+ :param watch: If True, input file will be watched for new lines and indexed in real time
+
Example record:
{
"ip": "43.134.51.142",
@@ -119,10 +151,11 @@ class ElasticIndexer:
}
'''
+ count = 0
records = []
with open(file_path, 'r') as file:
- for line in file:
+ for line in (file := follow(file) if watch else file):
line = line.strip()
if not line or not line.startswith('{'):
@@ -144,27 +177,84 @@ class ElasticIndexer:
struct['service'] = port_info['service']['name']
if 'banner' in port_info['service']:
- banner = port_info['service']['banner']
- match = re.search(r'\(Ref\.Id: (.*?)\)', banner)
- if match:
- struct['ref_id'] = match.group(1)
- else:
- struct['banner'] = banner
+ banner = ' '.join(port_info['service']['banner'].split()) # Remove extra whitespace
+ if banner:
+ match = re.search(r'\(Ref\.Id: (.*?)\)', banner)
+ if match:
+ struct['ref_id'] = match.group(1)
+ else:
+ struct['banner'] = banner
if self.dry_run:
print(struct)
else:
struct = {'_index': self.es_index, '_source': struct}
records.append(struct)
-
- if len(records) >= batch_size:
- success, _ = helpers.bulk(self.es, records)
- logging.info(f'Successfully indexed {success} records to {self.es_index}')
+ count += 1
+ if len(records) >= chunk['batch']:
+ self.bulk_index(records, file_path, chunk, count)
records = []
if records:
- success, _ = helpers.bulk(self.es, records)
- logging.info(f'Successfully indexed {success} records to {self.es_index}')
+ self.bulk_index(records, file_path, chunk, count)
+
+
+ def bulk_index(self, documents: list, file_path: str, chunk: dict, count: int):
+ '''
+ Index a batch of documents to Elasticsearch.
+
+ :param documents: List of documents to index
+ :param file_path: Path to the file being indexed
+ :param count: Total number of records processed
+ '''
+ remaining_documents = documents
+
+ parallel_bulk_config = {
+ 'client': self.es,
+ 'chunk_size': chunk['size'],
+ 'max_chunk_bytes': chunk['max_size'] * 1024 * 1024, # MB
+ 'thread_count': chunk['threads'],
+ 'queue_size': 2
+ }
+
+ while remaining_documents:
+ failed_documents = []
+
+ try:
+ for success, response in helpers.parallel_bulk(actions=remaining_documents, **parallel_bulk_config):
+ if not success:
+ failed_documents.append(response)
+
+ if not failed_documents:
+ ingested = parallel_bulk_config['chunk_size'] * parallel_bulk_config['thread_count']
+ logging.info(f'Successfully indexed {ingested:,} ({count:,} processed) records to {self.es_index} from {file_path}')
+ break
+
+ else:
+ logging.warning(f'Failed to index {len(failed_documents):,} failed documents! Retrying...')
+ remaining_documents = failed_documents
+ except Exception as e:
+ logging.error(f'Failed to index documents! ({e})')
+ time.sleep(30)
+
+
+def follow(file) -> str:
+ '''
+ Generator function that yields new lines in a file in real time.
+
+ :param file: File object to read from
+ '''
+
+ file.seek(0,2) # Go to the end of the file
+
+ while True:
+ line = file.readline()
+
+ if not line:
+ time.sleep(0.1)
+ continue
+
+ yield line
def main():
@@ -175,8 +265,8 @@ def main():
# General arguments
parser.add_argument('--dry-run', action='store_true', help='Dry run (do not index records to Elasticsearch)')
- parser.add_argument('--batch_size', type=int, default=50000, help='Number of records to index in a batch')
-
+ parser.add_argument('--watch', action='store_true', help='Watch the input file for new lines and index them in real time')
+
# Elasticsearch arguments
parser.add_argument('--host', default='localhost', help='Elasticsearch host')
parser.add_argument('--port', type=int, default=9200, help='Elasticsearch port')
@@ -186,54 +276,93 @@ def main():
parser.add_argument('--self-signed', action='store_false', help='Elasticsearch is using self-signed certificates')
# Elasticsearch indexing arguments
- parser.add_argument('--index', default='zone-files', help='Elasticsearch index name')
+ parser.add_argument('--index', default='masscan-logs', help='Elasticsearch index name')
parser.add_argument('--shards', type=int, default=1, help='Number of shards for the index')
parser.add_argument('--replicas', type=int, default=1, help='Number of replicas for the index')
+ # Batch arguments (for fine tuning performance)
+ parser.add_argument('--batch-max', type=int, default=10, help='Maximum size in MB of a batch')
+ parser.add_argument('--batch-size', type=int, default=5000, help='Number of records to index in a batch')
+ parser.add_argument('--batch-threads', type=int, default=2, help='Number of threads to use when indexing in batches')
+ # NOTE: Using --batch-threads as 4 and --batch-size as 10000 means we will process 40,000 records per-node before indexing, so 3 nodes would process 120,000 records before indexing
+
+ # Elasticsearch retry arguments
+ parser.add_argument('--retries', type=int, default=10, help='Number of times to retry indexing a batch before failing')
+ parser.add_argument('--timeout', type=int, default=30, help='Number of seconds to wait before retrying a batch')
+
args = parser.parse_args()
if not os.path.exists(args.input_path):
raise FileNotFoundError(f'Input file {args.input_path} does not exist')
+ elif not os.path.isdir(args.input_path) and not os.path.isfile(args.input_path):
+ raise ValueError(f'Input path {args.input_path} is not a file or directory')
if not args.dry_run:
if args.batch_size < 1:
raise ValueError('Batch size must be greater than 0')
+
+ elif args.retries < 1:
+ raise ValueError('Number of retries must be greater than 0')
+
+ elif args.timeout < 5:
+ raise ValueError('Timeout must be greater than 4')
+
+ elif args.batch_max < 1:
+ raise ValueError('Batch max size must be greater than 0')
+
+ elif args.batch_threads < 1:
+ raise ValueError('Batch threads must be greater than 0')
- if not args.host:
+ elif not args.host:
raise ValueError('Missing required Elasticsearch argument: host')
- if not args.api_key and (not args.user or not args.password):
+ elif not args.api_key and (not args.user or not args.password):
raise ValueError('Missing required Elasticsearch argument: either user and password or apikey')
- if args.shards < 1:
+ elif args.shards < 1:
raise ValueError('Number of shards must be greater than 0')
- if args.replicas < 0:
+ elif args.replicas < 0:
raise ValueError('Number of replicas must be greater than 0')
- logging.info(f'Connecting to Elasticsearch at {args.host}:{args.port}...')
-
- edx = ElasticIndexer(args.host, args.port, args.user, args.password, args.api_key, args.index, args.dry_run, args.self_signed)
-
+ edx = ElasticIndexer(args.host, args.port, args.user, args.password, args.api_key, args.index, args.dry_run, args.self_signed, args.retries, args.timeout)
+
if not args.dry_run:
+ time.sleep(3) # Delay to allow time for sniffing to complete
+
+ nodes = edx.get_cluster_size()
+ logging.info(f'Connected to {nodes:,} Elasticsearch node(s)')
+
edx.create_index(args.shards, args.replicas) # Create the index if it does not exist
+ chunk = {
+ 'size': args.batch_size,
+ 'max_size': args.batch_max * 1024 * 1024,
+ 'threads': args.batch_threads
+ }
+
+ chunk['batch'] = nodes * (chunk['size'] * chunk['threads'])
+ else:
+ chunk = {} # Ugly hack to get this working...
+
if os.path.isfile(args.input_path):
logging.info(f'Processing file: {args.input_path}')
- edx.process_file(args.input_path, args.batch_size)
+ edx.process_file(args.input_path, args.watch, chunk)
elif os.path.isdir(args.input_path):
- logging.info(f'Processing files in directory: {args.input_path}')
+ count = 1
+ total = len(os.listdir(args.input_path))
+ logging.info(f'Processing {total:,} files in directory: {args.input_path}')
for file in sorted(os.listdir(args.input_path)):
file_path = os.path.join(args.input_path, file)
if os.path.isfile(file_path):
- logging.info(f'Processing file: {file_path}')
- edx.process_file(file_path, args.batch_size)
-
- else:
- raise ValueError(f'Input path {args.input_path} is not a file or directory')
+ logging.info(f'[{count:,}/{total:,}] Processing file: {file_path}')
+ edx.process_file(file_path, args.watch, chunk)
+ count += 1
+ else:
+ logging.warning(f'[{count:,}/{total:,}] Skipping non-file: {file_path}')
if __name__ == '__main__':
- main()
+ main()
\ No newline at end of file
diff --git a/ingestors/ingest_massdns.py b/ingestors/ingest_massdns.py
index 3b9badf..f6f073d 100644
--- a/ingestors/ingest_massdns.py
+++ b/ingestors/ingest_massdns.py
@@ -9,13 +9,13 @@
# python $HOME/massdns/scripts/ptr.py | massdns -r $HOME/massdns/nameservers.txt -t PTR --filter NOERROR -o J -w $HOME/massdns/ptr.json
import argparse
-import json
import logging
import os
import time
try:
from elasticsearch import Elasticsearch, helpers
+ import sniff_patch
except ImportError:
raise ImportError('Missing required \'elasticsearch\' library. (pip install elasticsearch)')
@@ -25,11 +25,11 @@ logging.basicConfig(level=logging.INFO, format='%(asctime)s - %(name)s - %(level
class ElasticIndexer:
- def __init__(self, es_host: str, es_port: str, es_user: str, es_password: str, es_api_key: str, es_index: str, dry_run: bool = False, self_signed: bool = False):
+ def __init__(self, es_host: str, es_port: int, es_user: str, es_password: str, es_api_key: str, es_index: str, dry_run: bool = False, self_signed: bool = False, retries: int = 10, timeout: int = 30):
'''
Initialize the Elastic Search indexer.
- :param es_host: Elasticsearch host
+ :param es_host: Elasticsearch host(s)
:param es_port: Elasticsearch port
:param es_user: Elasticsearch username
:param es_password: Elasticsearch password
@@ -37,6 +37,8 @@ class ElasticIndexer:
:param es_index: Elasticsearch index name
:param dry_run: If True, do not initialize Elasticsearch client
:param self_signed: If True, do not verify SSL certificates
+ :param retries: Number of times to retry indexing a batch before failing
+ :param timeout: Number of seconds to wait before retrying a batch
'''
self.dry_run = dry_run
@@ -44,11 +46,25 @@ class ElasticIndexer:
self.es_index = es_index
if not dry_run:
+ es_config = {
+ 'hosts': [f'{es_host}:{es_port}'],
+ 'verify_certs': self_signed,
+ 'ssl_show_warn': self_signed,
+ 'request_timeout': timeout,
+ 'max_retries': retries,
+ 'retry_on_timeout': True,
+ 'sniff_on_start': False,
+ 'sniff_on_node_failure': True,
+ 'min_delay_between_sniffing': 60
+ }
if es_api_key:
- self.es = Elasticsearch([f'{es_host}:{es_port}'], headers={'Authorization': f'ApiKey {es_api_key}'}, verify_certs=self_signed, ssl_show_warn=self_signed)
+ es_config['headers'] = {'Authorization': f'ApiKey {es_api_key}'}
else:
- self.es = Elasticsearch([f'{es_host}:{es_port}'], basic_auth=(es_user, es_password), verify_certs=self_signed, ssl_show_warn=self_signed)
+ es_config['basic_auth'] = (es_user, es_password)
+
+ #self.es = Elasticsearch(**es_config)
+ self.es = sniff_patch.init_elasticsearch(**es_config)
def create_index(self, shards: int = 1, replicas: int = 1):
@@ -61,13 +77,10 @@ class ElasticIndexer:
},
'mappings': {
'properties': {
- 'ip': { 'type': 'ip' },
- 'port': { 'type': 'integer' },
- 'proto': { 'type': 'keyword' },
- 'service': { 'type': 'keyword' },
- 'banner': { 'type': 'text', 'fields': { 'keyword': { 'type': 'keyword', 'ignore_above': 256 } } },
- 'ref_id': { 'type': 'keyword' },
- 'seen': { 'type': 'date' }
+ 'ip': { 'type': 'ip' },
+ 'name': { 'type': 'keyword' },
+ 'record': { 'type': 'text', 'fields': { 'keyword': { 'type': 'keyword', 'ignore_above': 256 } } },
+ 'seen': { 'type': 'date' }
}
}
}
@@ -84,7 +97,17 @@ class ElasticIndexer:
logging.warning(f'Index \'{self.es_index}\' already exists.')
- def process_file(self, file_path: str, batch_size: int):
+ def get_cluster_size(self) -> int:
+ '''Get the number of nodes in the Elasticsearch cluster.'''
+
+ cluster_stats = self.es.cluster.stats()
+ number_of_nodes = cluster_stats['nodes']['count']['total']
+
+ return number_of_nodes
+
+
+
+ def process_file(self, file_path: str, watch: bool = False, chunk: dict = {}):
'''
Read and index PTR records in batches to Elasticsearch, handling large volumes efficiently.
@@ -92,78 +115,122 @@ class ElasticIndexer:
:param batch_size: Number of records to process before indexing
Example PTR record:
+ 0.6.229.47.in-addr.arpa. PTR 047-229-006-000.res.spectrum.com.
+ 0.6.228.75.in-addr.arpa. PTR 0.sub-75-228-6.myvzw.com.
+ 0.6.207.73.in-addr.arpa. PTR c-73-207-6-0.hsd1.ga.comcast.net.
+ 0.6.212.173.in-addr.arpa. PTR 173-212-6-0.cpe.surry.net.
+ 0.6.201.133.in-addr.arpa. PTR flh2-133-201-6-0.tky.mesh.ad.jp.
+
+ Will be indexed as:
{
- "name":"0.0.50.73.in-addr.arpa.",
- "type":"PTR",
- "class":"IN",
- "status":"NOERROR",
- "rx_ts":1704595370817617348,
- "data": {
- "answers": [
- {"ttl":3600,"type":"PTR","class":"IN","name":"0.0.50.73.in-addr.arpa.","data":"c-73-50-0-0.hsd1.il.comcast.net."}
- ],
- "authorities": [
- {"ttl":86400,"type":"NS","class":"IN","name":"73.in-addr.arpa.","data":"dns102.comcast.net."},
- {"ttl":86400,"type":"NS","class":"IN","name":"73.in-addr.arpa.","data":"dns103.comcast.net."},
- {"ttl":86400,"type":"NS","class":"IN","name":"73.in-addr.arpa.","data":"dns105.comcast.net."},
- {"ttl":86400,"type":"NS","class":"IN","name":"73.in-addr.arpa.","data":"dns101.comcast.net."},
- {"ttl":86400,"type":"NS","class":"IN","name":"73.in-addr.arpa.","data":"dns104.comcast.net."}
- ],
- "additionals":[
- {"ttl":105542,"type":"A","class":"IN","name":"dns101.comcast.net.","data":"69.252.250.103"},
- {"ttl":105542,"type":"A","class":"IN","name":"dns102.comcast.net.","data":"68.87.85.132"},
- {"ttl":105542,"type":"AAAA","class":"IN","name":"dns104.comcast.net.","data":"2001:558:100a:5:68:87:68:244"},
- {"ttl":105542,"type":"AAAA","class":"IN","name":"dns105.comcast.net.","data":"2001:558:100e:5:68:87:72:244"}
- ]
- },
- "flags": ["rd","ra"],
- "resolver":"103.144.64.173:53",
- "proto":"UDP"
+ "ip": "47.229.6.0",
+ "record": "047-229-006-000.res.spectrum.com.",
+ "seen": "2021-06-30T18:31:00Z"
}
'''
+ count = 0
records = []
with open(file_path, 'r') as file:
- for line in file:
+ for line in (file := follow(file) if watch else file):
line = line.strip()
if not line:
continue
- record = json.loads(line.strip())
+ parts = line.split()
- # Should we keep CNAME records? Can an IP address have a CNAME?
- if record['type'] != 'PTR':
- logging.error(record)
- raise ValueError(f'Unsupported record type: {record["type"]}')
-
- if record['class'] != 'IN':
- logging.error(record)
- raise ValueError(f'Unsupported record class: {record["class"]}')
+ if len(parts) < 3:
+ raise ValueError(f'Invalid PTR record: {line}')
+
+ name, record_type, data = parts[0].rstrip('.'), parts[1], ' '.join(parts[2:]).rstrip('.')
- if record['status'] != 'NOERROR':
- logging.warning(f'Skipping bad record: {record}')
+ if record_type != 'PTR':
continue
- record['ip'] = '.'.join(record['name'].replace('.in-addr.arpa', '').split('.')[::-1])
- record['name'] = record['name'].rstrip('.')
- record['timestamp'] = time.strftime("%Y-%m-%dT%H:%M:%SZ", time.gmtime(record.pop('rx_ts') / 1_000_000_000))
+ # Let's not index the PTR record if it's the same as the in-addr.arpa domain
+ if data == name:
+ continue
+
+ ip = '.'.join(name.replace('.in-addr.arpa', '').split('.')[::-1])
+
+ source = {
+ 'ip': ip,
+ 'record': data,
+ 'seen': time.strftime('%Y-%m-%dT%H:%M:%SZ', time.gmtime())
+ }
if self.dry_run:
- print(record)
+ print(source)
else:
- struct = {'_index': self.es_index, '_source': struct}
+ struct = {'_index': self.es_index, '_source': source}
records.append(struct)
-
- if len(records) >= batch_size:
- success, _ = helpers.bulk(self.es, records)
- logging.info(f'Successfully indexed {success} records to {self.es_index}')
+ count += 1
+ if len(records) >= chunk['batch']:
+ self.bulk_index(records, file_path, chunk, count)
records = []
if records:
- success, _ = helpers.bulk(self.es, records)
- logging.info(f'Successfully indexed {success} records to {self.es_index}')
+ self.bulk_index(records, file_path, chunk, count)
+
+
+ def bulk_index(self, documents: list, file_path: str, chunk: dict, count: int):
+ '''
+ Index a batch of documents to Elasticsearch.
+
+ :param documents: List of documents to index
+ :param file_path: Path to the file being indexed
+ :param count: Total number of records processed
+ '''
+ remaining_documents = documents
+
+ parallel_bulk_config = {
+ 'client': self.es,
+ 'chunk_size': chunk['size'],
+ 'max_chunk_bytes': chunk['max_size'] * 1024 * 1024, # MB
+ 'thread_count': chunk['threads'],
+ 'queue_size': 2
+ }
+
+ while remaining_documents:
+ failed_documents = []
+
+ try:
+ for success, response in helpers.parallel_bulk(actions=remaining_documents, **parallel_bulk_config):
+ if not success:
+ failed_documents.append(response)
+
+ if not failed_documents:
+ ingested = parallel_bulk_config['chunk_size'] * parallel_bulk_config['thread_count']
+ logging.info(f'Successfully indexed {ingested:,} ({count:,} processed) records to {self.es_index} from {file_path}')
+ break
+
+ else:
+ logging.warning(f'Failed to index {len(failed_documents):,} failed documents! Retrying...')
+ remaining_documents = failed_documents
+ except Exception as e:
+ logging.error(f'Failed to index documents! ({e})')
+ time.sleep(30)
+
+
+def follow(file) -> str:
+ '''
+ Generator function that yields new lines in a file in real time.
+
+ :param file: File object to read from
+ '''
+
+ file.seek(0,2) # Go to the end of the file
+
+ while True:
+ line = file.readline()
+
+ if not line:
+ time.sleep(0.1)
+ continue
+
+ yield line
def main():
@@ -174,63 +241,102 @@ def main():
# General arguments
parser.add_argument('--dry-run', action='store_true', help='Dry run (do not index records to Elasticsearch)')
- parser.add_argument('--batch_size', type=int, default=50000, help='Number of records to index in a batch')
-
- # Elasticsearch connection arguments
+ parser.add_argument('--watch', action='store_true', help='Watch the input file for new lines and index them in real time')
+
+ # Elasticsearch arguments
parser.add_argument('--host', default='localhost', help='Elasticsearch host')
parser.add_argument('--port', type=int, default=9200, help='Elasticsearch port')
parser.add_argument('--user', default='elastic', help='Elasticsearch username')
parser.add_argument('--password', default=os.getenv('ES_PASSWORD'), help='Elasticsearch password (if not provided, check environment variable ES_PASSWORD)')
- parser.add_argument('--api-key', help='Elasticsearch API Key for authentication')
+ parser.add_argument('--api-key', help='Elasticsearch API Key for authentication')
parser.add_argument('--self-signed', action='store_false', help='Elasticsearch is using self-signed certificates')
# Elasticsearch indexing arguments
- parser.add_argument('--index', default='zone-files', help='Elasticsearch index name')
+ parser.add_argument('--index', default='ptr-records', help='Elasticsearch index name')
parser.add_argument('--shards', type=int, default=1, help='Number of shards for the index')
parser.add_argument('--replicas', type=int, default=1, help='Number of replicas for the index')
+ # Batch arguments (for fine tuning performance)
+ parser.add_argument('--batch-max', type=int, default=10, help='Maximum size in MB of a batch')
+ parser.add_argument('--batch-size', type=int, default=5000, help='Number of records to index in a batch')
+ parser.add_argument('--batch-threads', type=int, default=2, help='Number of threads to use when indexing in batches')
+ # NOTE: Using --batch-threads as 4 and --batch-size as 10000 means we will process 40,000 records per-node before indexing, so 3 nodes would process 120,000 records before indexing
+
+ # Elasticsearch retry arguments
+ parser.add_argument('--retries', type=int, default=10, help='Number of times to retry indexing a batch before failing')
+ parser.add_argument('--timeout', type=int, default=30, help='Number of seconds to wait before retrying a batch')
+
args = parser.parse_args()
if not os.path.exists(args.input_path):
raise FileNotFoundError(f'Input file {args.input_path} does not exist')
+ elif not os.path.isdir(args.input_path) and not os.path.isfile(args.input_path):
+ raise ValueError(f'Input path {args.input_path} is not a file or directory')
if not args.dry_run:
if args.batch_size < 1:
raise ValueError('Batch size must be greater than 0')
+
+ elif args.retries < 1:
+ raise ValueError('Number of retries must be greater than 0')
+
+ elif args.timeout < 5:
+ raise ValueError('Timeout must be greater than 4')
+
+ elif args.batch_max < 1:
+ raise ValueError('Batch max size must be greater than 0')
+
+ elif args.batch_threads < 1:
+ raise ValueError('Batch threads must be greater than 0')
- if not args.host:
+ elif not args.host:
raise ValueError('Missing required Elasticsearch argument: host')
-
- if not args.api_key and (not args.user or not args.password):
+
+ elif not args.api_key and (not args.user or not args.password):
raise ValueError('Missing required Elasticsearch argument: either user and password or apikey')
-
- if args.shards < 1:
+
+ elif args.shards < 1:
raise ValueError('Number of shards must be greater than 0')
-
- if args.replicas < 1:
+
+ elif args.replicas < 0:
raise ValueError('Number of replicas must be greater than 0')
-
- logging.info(f'Connecting to Elasticsearch at {args.host}:{args.port}...')
-
- edx = ElasticIndexer(args.host, args.port, args.user, args.password, args.api_key, args.index, args.dry_run, args.self_signed)
+ edx = ElasticIndexer(args.host, args.port, args.user, args.password, args.api_key, args.index, args.dry_run, args.self_signed, args.retries, args.timeout)
+
if not args.dry_run:
+ time.sleep(5) # Delay to allow time for sniffing to complete
+
+ nodes = edx.get_cluster_size()
+ logging.info(f'Connected to {nodes:,} Elasticsearch node(s)')
+
edx.create_index(args.shards, args.replicas) # Create the index if it does not exist
+ chunk = {
+ 'size': args.batch_size,
+ 'max_size': args.batch_max * 1024 * 1024,
+ 'threads': args.batch_threads
+ }
+
+ chunk['batch'] = nodes * (chunk['size'] * chunk['threads'])
+ else:
+ chunk = {} # Ugly hack to get this working...
+
if os.path.isfile(args.input_path):
logging.info(f'Processing file: {args.input_path}')
- edx.process_file(args.input_path, args.batch_size)
+ edx.process_file(args.input_path, args.watch, chunk)
elif os.path.isdir(args.input_path):
- logging.info(f'Processing files in directory: {args.input_path}')
+ count = 1
+ total = len(os.listdir(args.input_path))
+ logging.info(f'Processing {total:,} files in directory: {args.input_path}')
for file in sorted(os.listdir(args.input_path)):
file_path = os.path.join(args.input_path, file)
if os.path.isfile(file_path):
- logging.info(f'Processing file: {file_path}')
- edx.process_file(file_path, args.batch_size)
-
- else:
- raise ValueError(f'Input path {args.input_path} is not a file or directory')
+ logging.info(f'[{count:,}/{total:,}] Processing file: {file_path}')
+ edx.process_file(file_path, args.watch, chunk)
+ count += 1
+ else:
+ logging.warning(f'[{count:,}/{total:,}] Skipping non-file: {file_path}')
diff --git a/ingestors/ingest_zone.py b/ingestors/ingest_zone.py
index d2634d1..9bc92fa 100644
--- a/ingestors/ingest_zone.py
+++ b/ingestors/ingest_zone.py
@@ -12,13 +12,21 @@
# Anomaly detection is in place to alert the user of any unexpected records.
+# WARNING NOTICE:
+#
+# The following zones need reindex due to previous unsupported record types:
+# - .wang (Contains a completely invalid line) (tjjm6hs65gL9KUFU76J747MB NS)
+# - .tel (NAPTR records were missing)
+# - .nu (RP records were missing
+# - .se (RP records were missing)
+
import argparse
import logging
import os
import time
try:
- from elasticsearch import Elasticsearch, helpers
+ from elasticsearch import Elasticsearch, helpers, ConnectionError, TransportError
except ImportError:
raise ImportError('Missing required \'elasticsearch\' library. (pip install elasticsearch)')
@@ -27,7 +35,7 @@ except ImportError:
logging.basicConfig(level=logging.INFO, format='%(asctime)s - %(name)s - %(levelname)s - %(message)s', datefmt='%m/%d %I:%M:%S')
# Record types to index
-record_types = ('a','aaaa','caa','cdnskey','cds','cname','dnskey','ds','mx','ns','nsec','nsec3','nsec3param','ptr','rrsig','sshfp','soa','srv','txt','type65534')
+record_types = ('a','aaaa','caa','cdnskey','cds','cname','dnskey','ds','mx','naptr','ns','nsec','nsec3','nsec3param','ptr','rrsig','rp','sshfp','soa','srv','txt','type65534')
class ElasticIndexer:
@@ -185,9 +193,16 @@ class ElasticIndexer:
records.append(struct)
count += 1
if len(records) >= batch_size:
- success, _ = helpers.bulk(self.es, records)
- logging.info(f'Successfully indexed {success:,} ({count:,}) records to {self.es_index} from {file_path}')
- records = []
+ while True:
+ try:
+ success, _ = helpers.bulk(self.es, records)
+ except (ConnectionError, TransportError) as e:
+ logging.error(f'Failed to index records to Elasticsearch. ({e})')
+ time.sleep(60)
+ else:
+ logging.info(f'Successfully indexed {success:,} ({count:,}) records to {self.es_index} from {file_path}')
+ records = []
+ break
last_domain = domain
@@ -199,8 +214,15 @@ class ElasticIndexer:
domain_records[domain][record_type].append({'ttl': ttl, 'data': data})
if records:
- success, _ = helpers.bulk(self.es, records)
- logging.info(f'Successfully indexed {success:,} ({count:,}) records to {self.es_index} from {file_path}')
+ while True:
+ try:
+ success, _ = helpers.bulk(self.es, records)
+ except (ConnectionError, TransportError) as e:
+ logging.error(f'Failed to index records to Elasticsearch. ({e})')
+ time.sleep(60)
+ else:
+ logging.info(f'Successfully indexed {success:,} ({count:,}) records to {self.es_index} from {file_path}')
+ break
def main():
@@ -222,7 +244,7 @@ def main():
parser.add_argument('--self-signed', action='store_false', help='Elasticsearch is using self-signed certificates')
# Elasticsearch indexing arguments
- parser.add_argument('--index', default='zone-files', help='Elasticsearch index name')
+ parser.add_argument('--index', default='dns-zones', help='Elasticsearch index name')
parser.add_argument('--shards', type=int, default=1, help='Number of shards for the index') # This depends on your cluster configuration
parser.add_argument('--replicas', type=int, default=1, help='Number of replicas for the index') # This depends on your cluster configuration
@@ -270,6 +292,5 @@ def main():
raise ValueError(f'Input path {args.input_path} is not a file or directory')
-
if __name__ == '__main__':
main()
\ No newline at end of file
diff --git a/ingestors/sniff_patch.py b/ingestors/sniff_patch.py
new file mode 100644
index 0000000..3d75402
--- /dev/null
+++ b/ingestors/sniff_patch.py
@@ -0,0 +1,81 @@
+# sniff_* patch for elastic 8 clients
+# Call init_elasticsearch() with normal Elasticsearch params
+# Only needed if you use sniff_* options and only works with basic auth, feel free to edit to your needs.
+
+import base64
+import elasticsearch._sync.client as client
+from elasticsearch.exceptions import SerializationError, ConnectionError
+
+
+def init_elasticsearch(*args, **kwargs):
+ client.default_sniff_callback = _override_sniff_callback(kwargs['basic_auth'])
+ return client.Elasticsearch(*args, **kwargs)
+
+
+def _override_sniff_callback(basic_auth):
+ """
+ Taken from https://github.com/elastic/elasticsearch-py/blob/8.8/elasticsearch/_sync/client/_base.py#L166
+ Completely unmodified except for adding the auth header to the elastic request.
+ Allows us to continue using the sniff_* options while this is broken in the library.
+
+ TODO: Remove this when this issue is patched:
+ https://github.com/elastic/elasticsearch-py/issues/2005
+ """
+ auth_str = base64.b64encode(':'.join(basic_auth).encode()).decode()
+ sniffed_node_callback = client._base._default_sniffed_node_callback
+
+ def modified_sniff_callback(transport, sniff_options):
+ for _ in transport.node_pool.all():
+ try:
+ meta, node_infos = transport.perform_request(
+ "GET",
+ "/_nodes/_all/http",
+ headers={
+ "accept": "application/vnd.elasticsearch+json; compatible-with=8",
+ # This auth header is missing in 8.x releases of the client, and causes 401s
+ "authorization": f"Basic {auth_str}"
+ },
+ request_timeout=(
+ sniff_options.sniff_timeout
+ if not sniff_options.is_initial_sniff
+ else None
+ ),
+ )
+ except (SerializationError, ConnectionError):
+ continue
+
+ if not 200 <= meta.status <= 299:
+ continue
+
+ node_configs = []
+ for node_info in node_infos.get("nodes", {}).values():
+ address = node_info.get("http", {}).get("publish_address")
+ if not address or ":" not in address:
+ continue
+
+ if "/" in address:
+ # Support 7.x host/ip:port behavior where http.publish_host has been set.
+ fqdn, ipaddress = address.split("/", 1)
+ host = fqdn
+ _, port_str = ipaddress.rsplit(":", 1)
+ port = int(port_str)
+ else:
+ host, port_str = address.rsplit(":", 1)
+ port = int(port_str)
+
+ assert sniffed_node_callback is not None
+ sniffed_node = sniffed_node_callback(
+ node_info, meta.node.replace(host=host, port=port)
+ )
+ if sniffed_node is None:
+ continue
+
+ # Use the node which was able to make the request as a base.
+ node_configs.append(sniffed_node)
+
+ if node_configs:
+ return node_configs
+
+ return []
+
+ return modified_sniff_callback
\ No newline at end of file