Backend API. Server-side code: this runs in a Codename One backend, not in the app on the device.
public final class JdbcRateLimiter
- Object
- JdbcRateLimiter
ImplementsRateLimiter
Counts in the server’s database, in the cn1_rate_limit table of
SecuritySchema, so that every process of a
deployment shares one limit.
http.rateLimit("/login", RateLimitKeys.clientAddress(),
new JdbcRateLimiter(dataSource, "login", 5, 60));
A fixed window: permits requests in each periodSeconds, counted from the
first request of the window. Every request is decided by one conditional
UPDATE and the number of rows it changed, so two processes counting at
the same moment cannot both take the last permit.
It costs a statement or two per request it counts, which is the price of
agreeing across processes; InMemoryRateLimiter costs none and counts for
its own process alone. A database that cannot be reached refuses nobody:
the limit is a courtesy to the server, and a store that is down must not
take the application with it.
Constructors
public JdbcRateLimiter(DataSource dataSource, String name, int permits, long periodSeconds) |
Methods
public RateLimiter derive(String name, int permits, long periodSeconds) | A limiter over the same table, under name, read on the same clock. |
public void setClock(Clock clock) | |
public boolean tryAcquire(String key) | Counts one request under key. |
public void reset(String key) | Deletes the row of key. |
public long retryAfterSeconds(String key) | How many seconds a request just refused under key should wait before trying again: what its Retry-After says. |
public int deleteExpired(long olderThanSeconds)
throws IOException | Deletes the counts of every limiter whose window ended more than olderThanSeconds ago; for a scheduled job, since a key that is never seen again leaves its row behind. |
Inherited methods
Constructor details
JdbcRateLimiter
public JdbcRateLimiter(DataSource dataSource, String name, int permits, long periodSeconds)Parameters
dataSourceDataSource- Not documented.
nameString- what keeps this limiter’s counts apart from another’s in the same table; two limiters of one name share their counts on purpose
permitsint- Not documented.
periodSecondslong- Not documented.
Method details
derive
public RateLimiter derive(String name, int permits, long periodSeconds)name, read on the same clock.Parameters
nameString- what keeps the new limiter’s counts apart
permitsint- how many requests a key may make in a period
periodSecondslong- the length of the period
setClock
public void setClock(Clock clock)tryAcquire
public boolean tryAcquire(String key)key.Returns
reset
public void reset(String key)key. A database that cannot be reached leaves the
count as it is, which is the safe way round for a bound on guesses.retryAfterSeconds
public long retryAfterSeconds(String key)key should wait before
trying again: what its Retry-After says. One second unless the limiter
knows better.deleteExpired
public int deleteExpired(long olderThanSeconds)
throws IOExceptionolderThanSeconds ago; for a scheduled job, since a key that is never
seen again leaves its row behind.