token_pools

Documentation

Index

• Constants
• type TokenPoolStorageBackend
  • func NewTokenPoolStorageBackend(name string, rawConfig interface{}) (TokenPoolStorageBackend, error)

Constants

const ResetInterval = time.Hour

how often tokens are reset

Types

type TokenPoolStorageBackend

type TokenPoolStorageBackend interface {
	// EnsureTokensAre will be called at init time, to ensure that the storage is in sync with the tokens listed in
	// the configuration.
	// For each given token, this should ensure that the token is present in the storage, and if not,
	// add it and assume it has its full hourly quota available and untouched.
	// Conversely, any token present in storage but not in the config should be deleted.
	EnsureTokensAre(tokens ...*types.TokenSpec) error

	// CheckOutToken checks out a token with its full hourly quota available, if there are any.
	// If there are no token available that hasn't been checked out yet, it should check out the token
	// that has the most calls remaining, if any.
	// The storage should keep track of when the token was checked out, and mark it as checked out
	// (if it was already checked out, the checked-out timestamp should be kept to its original value).
	// Errors are only for real storage errors; if the storage is working as intended, but simply
	// has no remaining token, it should just reply (nil, nil).
	CheckOutToken() (*types.Token, error)

	// UpdateTokenUsage should update a given checked-out token's count of remaining calls.
	// Trying to update a token that's unknown to the storage shouldn't yield an error, as this
	// can happen if a token has previously been checked out, before being removed from the config.
	UpdateTokenUsage(token string, remaining int) error

	// UpdateTokenRateLimit should update a given token's rate limit, in case the config gave
	// a wrong value for it.
	// Same as UpdateTokenUsage, trying to update a token that's unknown to the storage shouldn't yield an error.
	UpdateTokenRateLimit(token string, rateLimit int) error

	// CheckInTokens should make the storage iterate over all checked out tokens, and if they've been checked out
	// for longer than ResetInterval + gracePeriod, then check them back in and assume their full hourly quota
	// is available again.
	CheckInTokens(gracePeriod time.Duration) error

	// EstimateTotalRemainingCalls should return an estimate of the total count of remaining calls across all tokens
	// currently present in storage.
	EstimateTotalRemainingCalls() (int, error)
}

a TokenPoolStorageBackend abstracts away storing tokens and their metadata Implementations can be local, or distributed to allow for several instances to use the same storage. All implementations must be thread-safe.

func NewTokenPoolStorageBackend

func NewTokenPoolStorageBackend(name string, rawConfig interface{}) (TokenPoolStorageBackend, error)