Skip to content

Instantly share code, notes, and snippets.

Created August 20, 2018 08:08
Show Gist options
  • Star 0 You must be signed in to star a gist
  • Fork 0 You must be signed in to fork a gist
  • Save naviat/877238284c79b6512e3bbd91d2c54439 to your computer and use it in GitHub Desktop.
Save naviat/877238284c79b6512e3bbd91d2c54439 to your computer and use it in GitHub Desktop.
#!/usr/bin/env python3
## File:
## Author : Hai Dam
## Description : List full gc count for all es nodes
## python --es_host --max_full_gc 300
## --
## Created : <2018-08-19>
## Updated: Time-stamp: <2018-08-20 13:30:43>
import sys
import argparse, socket
import requests, json
# curl "http://$es_ip:9200/_nodes/stats" \
# | jq '[ .nodes | to_entries | sort_by(.value.jvm.gc.collectors.old.collection_count) | .[] | { node:, full_gc_count: .value.jvm.gc.collectors.old.collection_count } ]'
def get_es_gc_count(es_host, es_port):
url = "http://%s:%s/_nodes/stats" % (es_host, es_port)
r = requests.get(url)
if r.status_code != 200: raise Exception("Fail to run REST API: %s. Content: %s" % (url, r.content))
content_json = json.loads(r.content)
nodes_dict = content_json["nodes"]
res = []
for key in nodes_dict:
res.append([nodes_dict[key]["name"], nodes_dict[key]["host"], nodes_dict[key]["jvm"]["gc"]["collectors"]["old"]["collection_count"]])
return sorted(res, key=lambda item: item[2], reverse=True)
def check_es_gc_count(es_host, es_port, max_full_gc):
l = get_es_gc_count(es_host, es_port)
failed_nodes = []
print("ES nodes full gc, sorted in a reverse order")
for [node_name, node_ip, gc_count] in l:
print("%s\t%s\t%s" % (node_name, node_ip, gc_count))
if int(gc_count) >= max_full_gc:
failed_nodes.append((node_name, node_ip))
if len(failed_nodes) != 0:
print("Error: below nodes have full gc more than %d:" % (max_full_gc))
for (node_name, node_ip) in failed_nodes:
print("%s\t%s" % (node_name, node_ip))
return False
return True
if __name__ == '__main__':
parser = argparse.ArgumentParser()
parser.add_argument('--es_host', required=False, \
help="server ip or hostname for elasticsearch instance. Default value is ip of eth0", type=str)
parser.add_argument('--es_port', default='9200', required=False, \
help="server port for elasticsearch instance", type=str)
parser.add_argument('--max_full_gc', default='300', required=False, type=int, \
help="If some nodes have full gc more than this, fail the test")
l = parser.parse_args()
es_host = l.es_host
# get ip of eth0, if es_host is not given
if es_host is None:
s = socket.socket(socket.AF_INET, socket.SOCK_DGRAM)
s.connect(("", 80))
es_host = s.getsockname()[0]
ret = True
ret = check_es_gc_count(es_host, l.es_port, l.max_full_gc)
except Exception as e:
print("Unexpected error:%s, %s" % (sys.exc_info()[0], e))
if ret is False: sys.exit(1)
## File: ends
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment