Hello, I’m running Overpass API in Kubernetes with an EFS volume using the official image wiktorn/overpass-api:latest.
I use these environment variables:
env:
- name: OVERPASS_META
value: "yes"
- name: OVERPASS_MODE
value: "clone"
- name: OVERPASS_PLANET_URL
value: "https://planet.openstreetmap.org/planet/planet-latest.osm.bz2"
- name: OVERPASS_DIFF_URL
value: "https://planet.openstreetmap.org/replication/minute/"
- name: OVERPASS_RULES_LOAD
value: "10"
- name: OVERPASS_ALLOW_DUPLICATE_QUERIES
value: "yes"
And I set a 30GB, 7 cpu limits, and 500GB of disk space.
I use this locustfile to test it, and even with a single user, the server reaches the memory limit quickly, restarting in ungraceful pod restart.
from locust import HttpUser, task, between
import random
import base64
USERNAME = "admin"
PASSWORD = "admin"
QUERIES = [
# small bbox (most common)
"""
[out:json][timeout:25];
node["amenity"="restaurant"](36.85,10.15,36.86,10.16);
out;
""",
# medium bbox
"""
[out:json][timeout:25];
node["amenity"="restaurant"](36.84,10.14,36.87,10.18);
out;
""",
# heavier query (less frequent)
"""
[out:json][timeout:25];
node["shop"](36.8,10.1,36.9,10.2);
out;
"""
]
class OverpassUser(HttpUser):
wait_time = between(2, 10) # realistic think time
def on_start(self):
credentials = f"{USERNAME}:{PASSWORD}"
token = base64.b64encode(credentials.encode()).decode()
self.headers = {
"Authorization": f"Basic {token}",
"Content-Type": "application/x-www-form-urlencoded"
}
@task
def query_overpass(self):
query = random.choice(QUERIES)
self.client.post(
"/api/interpreter",
data={"data": query},
headers=self.headers
)
This is what the resource consumption looks like:
The pod then restarts and this issue runtime error: open64: 2 No such file or directory /db/db//osm3s_osm_base Unix_Socket::7 appears causing infinite restarts.
Is there a way to prevent the Overpass API server from reaching the limit and crashing?