Consider using isolation_level="IMMEDIATE" for write connections

[simonw]

We currently use the default isolation level for both read and write connections.

https://kerkour.com/sqlite-for-servers#use-immediate-transactions says:

This one may be one of the biggest footguns of SQLite.

By default, SQLite starts transactions in DEFERRED mode: they are considered read only. They are upgraded to a write transaction that requires a database lock in-flight, when query containing a write/update/delete statement is issued.

The problem is that by upgrading a transaction after it has started, SQLite will immediately return a SQLITE_BUSY error without respecting the busy_timeout previously mentioned, if the database is already locked by another connection.

This is why you should start your transactions with BEGIN IMMEDIATE instead of only BEGIN. If the database is locked when the transaction starts, SQLite will respect busy_timeout.

Hardest part is proving this issue one way or the other.

[simonw]

Here's the change:

diff --git a/datasette/database.py b/datasette/database.py
index ffe94ea7..e56dcf03 100644
--- a/datasette/database.py
+++ b/datasette/database.py
@@ -85,12 +85,17 @@ class Database:
             return "db"

     def connect(self, write=False):
+        kwargs = {
+            "uri": True,
+            "check_same_thread": False,
+        }
+        if write:
+            kwargs["isolation_level"] = "IMMEDIATE"
         if self.memory_name:
             uri = "file:{}?mode=memory&cache=shared".format(self.memory_name)
             conn = sqlite3.connect(
                 uri,
-                uri=True,
-                check_same_thread=False,
+                **kwargs,
             )
             if not write:
                 conn.execute("PRAGMA query_only=1")
@@ -110,9 +115,7 @@ class Database:
             qs = ""
         if self.mode is not None:
             qs = f"?mode={self.mode}"
-        conn = sqlite3.connect(
-            f"file:{self.path}{qs}", uri=True, check_same_thread=False
-        )
+        conn = sqlite3.connect(f"file:{self.path}{qs}", **kwargs)
         self._all_file_connections.append(conn)
         return conn

I don't want to commit this until I can run some kind of load test that demonstrates a problem with the default connection mode that is solved by switching to IMMEDIATE.

[simonw]

I tried running this locustfile.py using python -m locust -f locustfile.py and then simulating 100 users.

I ran sqlite-utils enable-wal data.db and then datasette data.db --root and used the /-/create-token page to create a token which I set as the DS_API_KEY environment variable.

from locust import HttpUser, task, between
import random
import json
import os

class APIUser(HttpUser):
    host = "http://localhost:8001"
    # wait_time = between(0.05, 0.1) # in seconds

    def __init__(self, *args, **kwargs):
        super().__init__(*args, **kwargs)
        self.id_counter = 0
        self.api_key = os.environ.get('DS_API_KEY')
        if not self.api_key:
            raise ValueError("DS_API_KEY environment variable is not set")
        self.headers = {
            'Content-Type': 'application/json',
            'Authorization': f'Bearer {self.api_key}'
        }

    @task(3)
    def get_data(self):
        self.client.get("/data", headers=self.headers)

    @task(3)
    def get_mytable(self):
        self.client.get("/data/mytable?_sort_desc=id", headers=self.headers)

    @task(2)
    def get_mytable_with_id(self):
        self.client.get("/data/mytable?id=5", headers=self.headers)

    @task(4)
    def post_create_data(self):
        self.id_counter += 1
        payload = {
            "table": "mytable",
            "rows": [
                {
                    "id": self.id_counter,
                    "name": random.choice(["Tarantula", "Black Widow", "Huntsman", "Wolf Spider", "Jumping Spider"])
                }
            ],
            "pk": "id",
            "replace": True
        }
        self.client.post("/data/-/create", data=json.dumps(payload), headers=self.headers)

On both my Mac and GitHub Codespaces I was unable to get it to trigger any database lock errors. Mac results:

I tried on Codespaces as well because I thought maybe my M2 performed too well and didn't trigger errors.

[simonw]

The currently open challenge here is this: create a locustfile.py which reliably triggers database locked errors when exercising a Datasette instance calling the JSON write API (and reading other endpoints at the same time).

[simonw]

Here's Codespaces Lotust run with WAL enabled (the macOS one above didn't have WAL enabled):

I used https://github.com/datasette/studio to run this