Glossary: logging, leaks and schema migrations — DedSec Project

Logging

Definition and boundaries

Recording contextual events with severity levels, timestamps and correlation IDs for diagnosis. Logs are observable output that can leak tokens or personal data if unsanitized.

Working miniature

import logging
logging.basicConfig(level=logging.INFO)
logging.info('processed %d records',3)

Guided investigation

Log a rejected record with a synthetic request ID while deliberately omitting credentials. Check which fields reach the handler, not only the console text.

Check your understanding

For Logging, change one input in the example and explain whether the observed behavior still satisfies the stated definition. Record the exact output and distinguish a program error from a mistaken assumption.

Memory leak

Definition and boundaries

Unexpected retention of objects past their useful lifetime. Long-lived caches, callbacks or references can cause this even with a garbage collector. High memory use from legitimate data is not necessarily a leak.

Working miniature

from collections import deque
recent=deque(maxlen=10)
for n in range(10000): recent.append(n)
assert len(recent)==10

Guided investigation

Compare an unbounded list with this bounded deque under repeated inputs. Use tracemalloc snapshots to identify retained allocations instead of guessing.

Retained references make objects live longer

An object may remain reachable through a growing registry even when its original local variable is deleted. Use a weak reference to observe this without exhausting memory.

import gc, weakref
class Entry: pass
registry = []
entry = Entry()
watch = weakref.ref(entry)
registry.append(entry)
del entry
gc.collect()
assert watch() is not None
registry.clear()
gc.collect()
assert watch() is None

Design a cache with a 100-entry maximum, then compare object retention after thousands of unique keys. Resident-set size alone is not sufficient to prove a leak because allocators may keep free pools; inspect surviving objects and owners.

Migration

Definition and boundaries

A controlled upgrade from one data schema or service version to another. Migrations must consider existing rows, partial deployment and rollback options rather than editing a schema once.

Working miniature

CREATE TABLE tasks(id INTEGER PRIMARY KEY,title TEXT);
ALTER TABLE tasks ADD COLUMN archived INTEGER NOT NULL DEFAULT 0;

Guided investigation

Use a disposable SQLite database containing older rows; verify that the new archived field is 0 and plan how a rollback would handle it.

Upgrade old rows without losing their values

A schema migration is successful only if existing records remain valid and the new schema is compatible with the application.

import sqlite3
db = sqlite3.connect(":memory:")
db.execute("CREATE TABLE tasks(id INTEGER PRIMARY KEY,title TEXT)")
db.execute("INSERT INTO tasks(title) VALUES (?)", ("review",))
with db:
    db.execute("ALTER TABLE tasks ADD COLUMN archived INTEGER NOT NULL DEFAULT 0")
assert db.execute("SELECT title,archived FROM tasks").fetchone() == ("review",0)
db.close()

Extend the lab with PRAGMA user_version and execute the upgrade twice. The second pass should detect that the new version is already installed. For destructive migrations, plan backups and rollback before applying changes; DDL transaction semantics vary across engines.