THIS IS A TEST INSTANCE. ALL YOUR CHANGES WILL BE LOST!!!!
Status
Current state: Under Discussion
Discussion thread: https://mail-archives.apache.org/mod_mbox/flink-dev/201902.mbox/browser
JIRA:
Released: <Flink Version>
Please keep the discussion on the mailing list rather than commenting on the wiki (wiki discussions get unwieldy fast).
Motivation
Flink has defined a few standard metrics for jobs, tasks and operators. It also supports custom metrics in various scenarios. However, so far there is no standard or conventional metric definition for the connectors. Each connector defines their own metrics at the moment. This complicates operation and monitoring. Admittedly, different connectors may have different metrics, but some commonly used metrics can probably be standardized. This FLIP proposes a set of standard connector metrics that each connector should emit if applicable.
In the future, the other projects in Flink ecosystem may rely on this metric convention. Therefore, the connector implementations are expected to follow the conventions when reporting the metrics.
Public Interfaces
We propose to introduce a set of conventional / standard metrics for the connectors.
It is important to mention that
- A connector implementation does not have to report all the defined metrics. But if a connector reports a metric of the same semantic defined below, the implementation should follow the convention.
- The following metric convention is not a complete list. More conventional metric will be added over time.
- The histogram metrics are usually very expensive. Due to its performance impact, we intentionally excluded them in this FLIP. Please see future work section for more details.
Source Metrics
Name | Type | Unit | Description |
numBytesIn | Counter | Bytes | The total number of input bytes since the source started |
numBytesInPerSecond | Meter | Bytes/Sec | The input bytes per second |
numRecordsIn | Counter | Records | (Existing operator metric) The total number of input records since the source started |
numRecordsInPerSecond | Meter | Records/Sec | (Existing operator metric) The input records per second |
| numRecordsInErrors | Counter | Records | The total number of record that failed to consume |
| currentFetchLatency | Gauge | ms | The latency occurred before Flink fetched the record. This metric is an instantaneous value recorded for the last processed record. This metric is provided because latency histogram could be expensive. The instantaneous latency value is usually a good enough indication of the latency. fetchLatency = FetchTime - EventTime |
| currentLatency | Gauge | ms | The latency occurred before the record is emitted by the source connector. This metric is an instantaneous value recorded for the last processed record. This metric is provided because latency histogram could be expensive. The instantaneous latency value is usually a good enough indication of the latency. latency = EmitTime - EventTime |
idleTime | Gauge | ms | The time in milliseconds that the source has not processed any record. idleTime = CurrentTime - LastRecordProcessTime |
| pendingBytes | Gauge | Bytes | The number of bytes that have not been fetched by the source. e.g. the remaining bytes in a file after the file descriptor reading position. |
| pendingRecords | Gauge | Records | The number of records that have not been fetched by the source. e.g. the available records after the consumer offset in a Kafka partition. |
Sink Metrics
Name | Type | Unit | Description |
numBytesOut | Counter | Bytes | The total number of output bytes since the source started |
numBytesOutPerSecond | Meter | Bytes/Sec | The output bytes per second |
numRecordsOut | Counter | Records | (Existing operator metric) The total number of output records since the source started |
numRecordsOutPerSecond | Meter | Records/Sec | (Existing operator metric) The output records per second |
| numRecordsOutErrors | Counter | Record | The total number of records failed to send |
| currentSendTime | Gauge | ms | The time it takes to send the last record. This metric is an instantaneous value recorded for the last processed record. |
Scope
The metric group for each source and sink would be the same as ordinary operator scope, i.e. default to <host>.taskmanager.<tm_id>.<job_name>.<operator_name>.<subtask_index>
Additional connector specific metrics should also use the same scope.
Native Connector Metrics
If the connector has its original metrics, the original metric names should still be kept, even some of the original metrics are exposed with standard metric names.
Future Work
Opt in/out metrics
In this FLIP, we intentionally left some of the useful but potentially expensive metrics out of the scope. For example:
Name | Type | Unit | Description |
recordSize | Histogram | Bytes | The size of a record. |
fetchLatency | Histogram | ms | The latency occurred before Flink fetched the record. fetchLatency = FetchTime - EventTime |
latency | Histogram | ms | The latency occurred before the record is emitted by the source connector. latency = EmitTime - EventTime |
sendTime | Histogram | ms | The time it takes to send a record |
We plan to add these metrics to the convention by introducing optional metrics to allow user opt in/out these expensive metrics on demand. This will be discussed in a separate FLIP.
Proposed Changes
- Add the proposed metrics to the existing connectors.
- Mark the old metrics as deprecated if necessary.
- Correct the scope and metric names of the connectors if needed.
Compatibility, Deprecation, and Migration Plan
This FLIP proposes adding new metrics to the connectors and mark the old duplicate metrics as deprecated if necessary.
After some releases, we would like to remove some of the old metrics that are duplicates of the new standard metrics. But there is no strict timeline for that.
Test Plan
Standard connector metric test suites will be created to ensure the connector names are implemented correctly.
Rejected Alternatives
None