apache/celeborn
1
0
Fork 0
Code Issues Pull Requests Actions 1 Packages Projects Releases Wiki Activity
a77a8eb8fd
celeborn/docs/monitoring.md
SteNicholas 92777c3ff2 [CELEBORN-927][DOC] Correct celeborn.metrics.conf.*.sink.csv.class configuration example for a CSV sink 
### What changes were proposed in this pull request?

Correct `celeborn.metrics.conf.*.sink.csv.class` configuration example for a CSV sink.

### Why are the changes needed?

`celeborn.metrics.conf.*.sink.csv.class` configuration example for a CSV sink is wrong, which value should be `org.apache.celeborn.common.metrics.sink.CsvSink`.

### Does this PR introduce _any_ user-facing change?

No.

### How was this patch tested?

None.

Closes #1865 from SteNicholas/CELEBORN-927.

Authored-by: SteNicholas <programgeek@163.com>
Signed-off-by: zky.zhoukeyong <zky.zhoukeyong@alibaba-inc.com>
2023-08-30 16:11:03 +08:00

318 lines
16 KiB
Markdown
Raw Blame History

This file contains ambiguous Unicode characters

This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

---
license: |
Licensed to the Apache Software Foundation (ASF) under one or more
contributor license agreements. See the NOTICE file distributed with
this work for additional information regarding copyright ownership.
The ASF licenses this file to You under the Apache License, Version 2.0
(the "License"); you may not use this file except in compliance with
the License. You may obtain a copy of the License at
https://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
---
Monitoring
===
There are two ways to monitor Celeborn cluster: Prometheus metrics and REST API.
## Metrics
Celeborn has a configurable metrics system based on the
[Dropwizard Metrics Library](http://metrics.dropwizard.io/4.2.0).
This allows users to report Celeborn metrics to a variety of sinks including HTTP, JMX, CSV
files and prometheus servlet. The metrics are generated by sources embedded in the Celeborn code base.
They provide instrumentation for specific activities and Celeborn components.
The metrics system is configured via a configuration file that Celeborn expects to be present
at `$CELEBORN_HOME/conf/metrics.properties`. A custom file location can be specified via the
`spark.metrics.conf` [configuration property](https://celeborn.apache.org/configuration/#metrics).
Instead of using the configuration file, a set of configuration parameters with prefix
`celeborn.metrics.conf.` can be used.
Celeborn's metrics are divided into two
_instances_ corresponding to Celeborn components. The following instances are currently supported:
* `master`: The Celeborn cluster master process.
* `worker`: The Celeborn cluster worker process.
Each instance can report to zero or more _sinks_. Sinks are contained in the
`org.apache.celeborn.common.metrics.sink` package:
* `CSVSink`: Exports metrics data to CSV files at regular intervals.
* `PrometheusServlet`: Adds a servlet within the existing Celeborn REST API to serve metrics data in Prometheus format.
* `GraphiteSink`: Sends metrics to a Graphite node.
The syntax of the metrics configuration file and the parameters available for each sink are defined
in an example configuration file,
`$CELEBORN_HOME/conf/metrics.properties.template`.
When using Celeborn configuration parameters instead of the metrics configuration file, the relevant
parameter names are composed by the prefix `celeborn.metrics.conf.` followed by the configuration
details, i.e. the parameters take the following form:
`celeborn.metrics.conf.[instance|*].sink.[sink_name].[parameter_name]`.
This example shows a list of Celeborn configuration parameters for a CSV sink:
```
"celeborn.metrics.conf.*.sink.csv.class"="org.apache.celeborn.common.metrics.sink.CsvSink"
"celeborn.metrics.conf.*.sink.csv.period"="1"
"celeborn.metrics.conf.*.sink.csv.unit"=minutes
"celeborn.metrics.conf.*.sink.csv.directory"=/tmp/
```
Default values of the Celeborn metrics configuration are as follows:
```
*.sink.prometheusServlet.class=org.apache.celeborn.common.metrics.sink.PrometheusServlet
```
Additional sources can be configured using the metrics configuration file or the configuration
parameter `spark.metrics.conf.[component_name].source.jvm.class=[source_name]`. At present the
no source is the available optional source. For example the following configuration parameter
activates the Example source:
`"celeborn.metrics.conf.*.source.jvm.class"="org.apache.celeborn.common.metrics.source.ExampleSource"`
### Available metrics providers
Metrics used by Celeborn are of multiple types: gauge, counter, histogram, meter and timer,
see [Dropwizard library documentation for details](https://metrics.dropwizard.io/4.2.0/getting-started.html).
The following list of components and metrics reports the name and some details about the available metrics,
grouped per component instance and source namespace.
The most common time of metrics used in Celeborn instrumentation are gauges and counters.
Counters can be recognized as they have the `.count` suffix. Timers, meters and histograms are annotated
in the list, the rest of the list elements are metrics of type gauge.
The large majority of metrics are active as soon as their parent component instance is configured,
some metrics require also to be enabled via an additional configuration parameter, the details are
reported in the list.
#### Master
These metrics are exposed by Celeborn master.
- namespace=master
- WorkerCount
- LostWorkers
- ExcludedWorkerCount
- RegisteredShuffleCount
- IsActiveMaster
- PartitionSize
- The size of estimated shuffle partition.
- OfferSlotsTime
- The time for masters to handle `RequestSlots` request when registering shuffle.
- namespace=CPU
- JVMCPUTime
- namespace=system
- LastMinuteSystemLoad
- The average system load for the last minute.
- AvailableProcessors
- namespace=JVM
- This source provides information on JVM metrics using the
[Dropwizard/Codahale Metric Sets for JVM instrumentation](https://metrics.dropwizard.io/4.2.0/manual/jvm.html)
and in particular the metric sets BufferPoolMetricSet, GarbageCollectorMetricSet and MemoryUsageGaugeSet.
- namespace=ResourceConsumption
- **notes:**
- This metrics data is generated for each user and they are identified using a metric tag.
- diskFileCount
- diskBytesWritten
- hdfsFileCount
- hdfsBytesWritten
#### Worker
These metrics are exposed by Celeborn worker.
- namespace=worker
- CommitFilesTime
- The time for a worker to flush buffers and close files related to specified shuffle.
- ReserveSlotsTime
- FlushDataTime
- The time for a worker to write a buffer which is 256KB by default to storage.
- OpenStreamTime
- The time for a worker to process openStream RPC and return StreamHandle.
- FetchChunkTime
- The time for a worker to fetch a chunk which is 8MB by default from a reduced partition.
- PrimaryPushDataTime
- The time for a worker to handle a pushData RPC sent from a celeborn client.
- ReplicaPushDataTime
- The time for a worker to handle a pushData RPC sent from a celeborn worker by replicating.
- WriteDataFailCount
- ReplicateDataFailCount
- ReplicateDataWriteFailCount
- ReplicateDataCreateConnectionFailCount
- ReplicateDataConnectionExceptionCount
- ReplicateDataTimeoutCount
- PushDataHandshakeFailCount
- RegionStartFailCount
- RegionFinishFailCount
- PrimaryPushDataHandshakeTime
- ReplicaPushDataHandshakeTime
- PrimaryRegionStartTime
- ReplicaRegionStartTime
- PrimaryRegionFinishTime
- ReplicaRegionFinishTime
- TakeBufferTime
- The time for a worker to take out a buffer from a disk flusher.
- RegisteredShuffleCount
- SlotsAllocated
- NettyMemory
- The total amount of off-heap memory used by celeborn worker.
- SortTime
- The time for a worker to sort a shuffle file.
- SortMemory
- The memory used by sorting shuffle files.
- SortingFiles
- SortedFiles
- SortedFileSize
- DiskBuffer
- The memory occupied by pushData and pushMergedData which should be written to disk.
- PausePushData
- The count for a worker to stop receiving pushData from clients because of back pressure.
- PausePushDataAndReplicate
- The count for a worker to stop receiving pushData from clients and other workers because of back pressure.
- BufferStreamReadBuffer
- The memory used by credit stream read buffer.
- ReadBufferDispatcherRequestsLength
- The queue size of read buffer allocation requests.
- ReadBufferAllocatedCount
- Allocated read buffer count.
- CreditStreamCount
- Stream count for map partition reading streams.
- ActiveMapPartitionCount
- DeviceOSFreeBytes
- DeviceOSTotalBytes
- DeviceCelebornFreeBytes
- DeviceCelebornTotalBytes
- PotentialConsumeSpeed
- UserProduceSpeed
- WorkerConsumeSpeed
- push_server_usedHeapMemory
- push_server_usedDirectMemory
- push_server_numAllocations
- push_server_numTinyAllocations
- push_server_numSmallAllocations
- push_server_numNormalAllocations
- push_server_numHugeAllocations
- push_server_numDeallocations
- push_server_numTinyDeallocations
- push_server_numSmallDeallocations
- push_server_numNormalDeallocations
- push_server_numHugeDeallocations
- push_server_numActiveAllocations
- push_server_numActiveTinyAllocations
- push_server_numActiveSmallAllocations
- push_server_numActiveNormalAllocations
- push_server_numActiveHugeAllocations
- push_server_numActiveBytes
- replicate_server_usedHeapMemory
- replicate_server_usedDirectMemory
- replicate_server_numAllocations
- replicate_server_numTinyAllocations
- replicate_server_numSmallAllocations
- replicate_server_numNormalAllocations
- replicate_server_numHugeAllocations
- replicate_server_numDeallocations
- replicate_server_numTinyDeallocations
- replicate_server_numSmallDeallocations
- replicate_server_numNormalDeallocations
- replicate_server_numHugeDeallocations
- replicate_server_numActiveAllocations
- replicate_server_numActiveTinyAllocations
- replicate_server_numActiveSmallAllocations
- replicate_server_numActiveNormalAllocations
- replicate_server_numActiveHugeAllocations
- replicate_server_numActiveBytes
- fetch_server_usedHeapMemory
- fetch_server_usedDirectMemory
- fetch_server_numAllocations
- fetch_server_numTinyAllocations
- fetch_server_numSmallAllocations
- fetch_server_numNormalAllocations
- fetch_server_numHugeAllocations
- fetch_server_numDeallocations
- fetch_server_numTinyDeallocations
- fetch_server_numSmallDeallocations
- fetch_server_numNormalDeallocations
- fetch_server_numHugeDeallocations
- fetch_server_numActiveAllocations
- fetch_server_numActiveTinyAllocations
- fetch_server_numActiveSmallAllocations
- fetch_server_numActiveNormalAllocations
- fetch_server_numActiveHugeAllocations
- fetch_server_numActiveBytes
- namespace=CPU
- JVMCPUTime
- namespace=system
- LastMinuteSystemLoad
- Returns the system load average for the last minute.
- AvailableProcessors
- namespace=JVM
- This source provides information on JVM metrics using the
[Dropwizard/Codahale Metric Sets for JVM instrumentation](https://metrics.dropwizard.io/4.2.0/manual/jvm.html)
and in particular the metric sets BufferPoolMetricSet, GarbageCollectorMetricSet and MemoryUsageGaugeSet.
**Note:**
The Netty DirectArenaMetrics named like `push/fetch/replicate_server_numXX` are not exposed by default, nor in Grafana dashboard.
If there is a need, you can enable `celeborn.network.memory.allocator.verbose.metric` to expose these metrics.
## REST API
In addition to viewing the metrics, Celeborn also support REST API. This gives developers
an easy way to create new visualizations and monitoring tools for Celeborn and
also easy for users to get the running status of the service. The REST API is available for
both master and worker. The endpoints are mounted at `host:port`. For example,
for the master, they would typically be accessible
at `http://<master-prometheus-host>:<master-prometheus-port><path>`, and
for the worker, at `http://<worker-prometheus-host>:<worker-prometheus-port><path>`.
The configuration of `<master-prometheus-host>`, `<master-prometheus-port>`, `<worker-prometheus-host>`, `<worker-prometheus-port>` as below:
| Key | Default | Description | Since |
|-----------------------------------------|---------|----------------------------|-------|
| celeborn.metrics.master.prometheus.host | 0.0.0.0 | Master's Prometheus host. | 0.2.0 |
| celeborn.metrics.master.prometheus.port | 9098 | Master's Prometheus port. | 0.2.0 |
| celeborn.metrics.worker.prometheus.host | 0.0.0.0 | Worker's Prometheus host. | 0.2.0 |
| celeborn.metrics.worker.prometheus.port | 9096 | Worker's Prometheus port. | 0.2.0 |
### Available API providers
API path listed as below:
#### Master
| Path | Meaning |
|-----------------------|-------------------------------------------------------------------------------------------------------------|
| /metrics/prometheus | List the metrics data in prometheus format of the master. |
| /conf | List the conf setting of the master. |
| /workerInfo | List worker information of the service. It will list all registered workers 's information. |
| /lostWorkers | List all lost workers of the master. |
| /excludedWorkers | List all excluded workers of the master. |
| /threadDump | List the current thread dump of the master. |
| /hostnames | List all running application's LifecycleManager's hostnames of the cluster. |
| /applications | List all running application's ids of the cluster. |
| /shuffles | List all running shuffle keys of the service. It will return all running shuffle's key of the cluster. |
| /listTopDiskUsedApps | List the top disk usage application ids. It will return the top disk usage application ids for the cluster. |
#### Worker
| Path | Meaning |
|----------------------------|----------------------------------------------------------------------------------------------------------|
| /metrics/prometheus | List the metrics data in prometheus format of the worker. |
| /conf | List the conf setting of the worker. |
| /workerInfo | List the worker information of the worker. |
| /threadDump | List the current thread dump of the worker. |
| /shuffles | List all the running shuffle keys of the worker. It only return keys of shuffles running in that worker. |
| /listTopDiskUsedApps | List the top disk usage application ids. It only return application ids running in that worker. |
| /listPartitionLocationInfo | List all the living PartitionLocation information in that worker. |
| /unavailablePeers | List the unavailable peers of the worker, this always means the worker connect to the peer failed. |
| /isShutdown | Show if the worker is during the process of shutdown. |
| /isRegistered | Show if the worker is registered to the master success. |
| /exit?type=${TYPE} | Trigger this worker to exit. Legal `type`s are 'DECOMMISSION‘, 'GRACEFUL' and 'IMMEDIATELY' |
Reference in New Issue View Git Blame Copy Permalink