Module: Rdkafka::Defaults

Defined in:

lib/rdkafka/defaults.rb

Overview

Provides default timeout and configuration values used throughout the library.

These constants standardize timing values across consumers, producers, and admin clients. Values are specified in milliseconds (ms) unless otherwise noted.

Constant Summary

CONSUMER_COMMITTED_TIMEOUT_MS =

Consumer timeout for fetching committed offsets

See Also:

• Consumer#committed

2_000

CONSUMER_QUERY_WATERMARK_TIMEOUT_MS =

Consumer timeout for querying watermark offsets

See Also:

• Consumer#query_watermark_offsets

1_000

CONSUMER_LAG_TIMEOUT_MS =

Consumer timeout for lag calculations

See Also:

• Consumer#lag

1_000

CONSUMER_OFFSETS_FOR_TIMES_TIMEOUT_MS =

Consumer timeout for offset-by-timestamp lookups

See Also:

• Consumer#offsets_for_times

1_000

CONSUMER_CLUSTER_ID_TIMEOUT_MS =

Consumer timeout for fetching the cluster id when it is not already cached from metadata

See Also:

• Consumer#cluster_id

1_000

CONSUMER_POLL_TIMEOUT_MS =

Consumer timeout for poll operations (used in each iteration)

See Also:

• Consumer#each

250

CONSUMER_SEEK_TIMEOUT_MS =

Consumer timeout for seek operations (0 = non-blocking)

See Also:

• Consumer#seek_by

0

CONSUMER_EVENTS_POLL_TIMEOUT_MS =

Consumer timeout for events_poll (0 = non-blocking async)

See Also:

• Consumer#events_poll

0

PRODUCER_FLUSH_TIMEOUT_MS =

Producer timeout for flush operations

See Also:

• Producer#flush

5_000

PRODUCER_PURGE_FLUSH_TIMEOUT_MS =

Producer timeout for flush during purge

See Also:

• Producer#purge

100

PRODUCER_PURGE_SLEEP_INTERVAL_MS =

Sleep interval used in producer purge loop

See Also:

• Producer#purge

1

PRODUCER_SEND_OFFSETS_TIMEOUT_MS =

Timeout for transactional send_offsets_to_transaction

See Also:

• Producer#send_offsets_to_transaction

5_000

METADATA_TIMEOUT_MS =

Default timeout for metadata requests

See Also:

• Metadata#initialize
• Helpers::Metadata#metadata

2_000

METADATA_MAX_RETRIES =

Hard ceiling on metadata fetch attempts on transient errors (backstop; the retry budget normally ends the loop first)

See Also:

• Metadata#initialize

10

METADATA_MIN_ATTEMPTS =

Minimum metadata fetch attempts before the retry budget may end the loop, so a slow broker (whose requests each consume the full timeout) still gets a few tries

See Also:

• Metadata#initialize

3

METADATA_RETRY_BUDGET_MS =

Soft wall-clock budget for the whole metadata retry loop; past it (and past METADATA_MIN_ATTEMPTS) the loop stops so a synchronous fetch cannot block the caller for long

See Also:

• Metadata#initialize

5_000

METADATA_RETRY_BACKOFF_BASE_MS =

Base backoff time for metadata retry (100ms = 0.1s)

See Also:

• Metadata#initialize

100

METADATA_RETRY_BACKOFF_MAX_MS =

Maximum backoff time between metadata retries. Caps the exponential backoff so a long retry sequence against an unhealthy cluster cannot block the calling thread for minutes.

See Also:

• Metadata#initialize

1_000

HANDLE_WAIT_TIMEOUT_MS =

Default wait timeout for operation handles

See Also:

• AbstractHandle#wait

60_000

NATIVE_KAFKA_POLL_TIMEOUT_MS =

Polling interval for NativeKafka background thread

See Also:

• NativeKafka#initialize
• Config#producer
• Config#admin

100

NATIVE_KAFKA_SYNCHRONIZE_SLEEP_INTERVAL_MS =

Sleep interval used in NativeKafka#synchronize wait loop

See Also:

• NativeKafka#synchronize

10

PARTITIONS_COUNT_CACHE_TTL_MS =

TTL for partitions count cache entries (30 seconds)

See Also:

• Producer::PartitionsCountCache

30_000