Skip to content

Commit

Permalink
Docs: Remove references to Zk-based segment loading (apache#16360)
Browse files Browse the repository at this point in the history
Follow up to apache#15705

Changes:
- Remove references to ZK-based segment loading in the docs
- Fix doc for existing config `druid.coordinator.loadqueuepeon.http.repeatDelay`
  • Loading branch information
kfaraz authored May 1, 2024
1 parent 834b0ed commit 51104e8
Show file tree
Hide file tree
Showing 3 changed files with 6 additions and 27 deletions.
12 changes: 2 additions & 10 deletions docs/configuration/index.md
Original file line number Diff line number Diff line change
Expand Up @@ -160,7 +160,6 @@ Druid interacts with ZooKeeper through a set of standard path configurations. We
|`druid.zk.paths.propertiesPath`|ZooKeeper properties path.|`${druid.zk.paths.base}/properties`|
|`druid.zk.paths.announcementsPath`|Druid service announcement path.|`${druid.zk.paths.base}/announcements`|
|`druid.zk.paths.liveSegmentsPath`|Current path for where Druid services announce their segments.|`${druid.zk.paths.base}/segments`|
|`druid.zk.paths.loadQueuePath`|Entries here cause Historical services to load and drop segments.|`${druid.zk.paths.base}/loadQueue`|
|`druid.zk.paths.coordinatorPath`|Used by the Coordinator for leader election.|`${druid.zk.paths.base}/coordinator`|
|`druid.zk.paths.servedSegmentsPath`|Deprecated. Legacy path for where Druid services announce their segments.|`${druid.zk.paths.base}/servedSegments`|

Expand Down Expand Up @@ -875,7 +874,8 @@ These Coordinator static configurations can be defined in the `coordinator/runti
|`druid.coordinator.kill.maxSegments`|The number of unused segments to kill per kill task. This number must be greater than 0. This only applies when `druid.coordinator.kill.on=true`.|100|
|`druid.coordinator.balancer.strategy`|Specify the type of balancing strategy for the Coordinator to use to distribute segments among the Historical services. `cachingCost` is logically equivalent to `cost` but is more CPU-efficient on large clusters. `diskNormalized` weights the costs according to the servers' disk usage ratios - there are known issues with this strategy distributing segments unevenly across the cluster. `random` distributes segments among services randomly.|`cost`|
|`druid.coordinator.balancer.cachingCost.awaitInitialization`|Whether to wait for segment view initialization before creating the `cachingCost` balancing strategy. This property is enabled only when `druid.coordinator.balancer.strategy` is `cachingCost`. If set to true, the Coordinator will not start to assign segments, until the segment view is initialized. If set to false, the Coordinator will fallback to use the `cost` balancing strategy only if the segment view is not initialized yet. It may take much time to wait for the initialization since the `cachingCost` balancing strategy involves much computing to build itself.|false|
|`druid.coordinator.loadqueuepeon.repeatDelay`|The start and repeat delay for the `loadqueuepeon`, which manages the load and drop of segments.|`PT0.050S` (50 ms)|
|`druid.coordinator.loadqueuepeon.http.repeatDelay`|The start and repeat delay (in milliseconds) for the load queue peon, which manages the load/drop queue of segments for any server.|1 minute|
|`druid.coordinator.loadqueuepeon.http.batchSize`|Number of segment load/drop requests to batch in one HTTP request. Note that it must be smaller than `druid.segmentCache.numLoadingThreads` config on Historical service.|1|
|`druid.coordinator.asOverlord.enabled`|Boolean value for whether this Coordinator service should act like an Overlord as well. This configuration allows users to simplify a Druid cluster by not having to deploy any standalone Overlord services. If set to true, then Overlord console is available at `http://coordinator-host:port/console.html` and be sure to set `druid.coordinator.asOverlord.overlordService` also.|false|
|`druid.coordinator.asOverlord.overlordService`| Required, if `druid.coordinator.asOverlord.enabled` is `true`. This must be same value as `druid.service` on standalone Overlord services and `druid.selectors.indexing.serviceName` on Middle Managers.|NULL|
|`druid.centralizedDatasourceSchema.enabled`|Boolean flag for enabling datasource schema building on the Coordinator. Note, when using MiddleManager to launch task, set `druid.indexer.fork.property.druid.centralizedDatasourceSchema.enabled` in MiddleManager runtime config. |false|
Expand Down Expand Up @@ -905,15 +905,8 @@ These Coordinator static configurations can be defined in the `coordinator/runti
|Property|Possible values|Description|Default|
|--------|---------------|-----------|-------|
|`druid.serverview.type`|batch or http|Segment discovery method to use. "http" enables discovering segments using HTTP instead of ZooKeeper.|http|
|`druid.coordinator.loadqueuepeon.type`|curator or http|Implementation to use to assign segment loads and drops to historicals. Curator-based implementation is now deprecated, so you should transition to using HTTP-based segment assignments.|http|
|`druid.coordinator.segment.awaitInitializationOnStart`|true or false|Whether the Coordinator will wait for its view of segments to fully initialize before starting up. If set to 'true', the Coordinator's HTTP server will not start up, and the Coordinator will not announce itself as available, until the server view is initialized.|true|

###### Additional config when "http" loadqueuepeon is used

|Property|Description|Default|
|--------|-----------|-------|
|`druid.coordinator.loadqueuepeon.http.batchSize`|Number of segment load/drop requests to batch in one HTTP request. Note that it must be smaller than `druid.segmentCache.numLoadingThreads` config on Historical service.|1|

##### Metadata retrieval

|Property|Description|Default|
Expand Down Expand Up @@ -1653,7 +1646,6 @@ These Historical configurations can be defined in the `historical/runtime.proper
|`druid.segmentCache.numLoadingThreads`|How many segments to drop or load concurrently from deep storage. Note that the work of loading segments involves downloading segments from deep storage, decompressing them and loading them to a memory mapped location. So the work is not all I/O Bound. Depending on CPU and network load, one could possibly increase this config to a higher value.|max(1,Number of cores / 6)|
|`druid.segmentCache.numBootstrapThreads`|How many segments to load concurrently during historical startup.|`druid.segmentCache.numLoadingThreads`|
|`druid.segmentCache.lazyLoadOnStart`|Whether or not to load segment columns metadata lazily during historical startup. When set to true, Historical startup time will be dramatically improved by deferring segment loading until the first time that segment takes part in a query, which will incur this cost instead.|false|
|`druid.coordinator.loadqueuepeon.curator.numCallbackThreads`|Number of threads for executing callback actions associated with loading or dropping of segments. One might want to increase this number when noticing clusters are lagging behind w.r.t. balancing segments across historical nodes.|2|
|`druid.segmentCache.numThreadsToLoadSegmentsIntoPageCacheOnDownload`|Number of threads to asynchronously read segment index files into null output stream on each new segment download after the Historical service finishes bootstrapping. Recommended to set to 1 or 2 or leave unspecified to disable. See also `druid.segmentCache.numThreadsToLoadSegmentsIntoPageCacheOnBootstrap`|0|
|`druid.segmentCache.numThreadsToLoadSegmentsIntoPageCacheOnBootstrap`|Number of threads to asynchronously read segment index files into null output stream during Historical service bootstrap. This thread pool is terminated after Historical service finishes bootstrapping. Recommended to set to half of available cores. If left unspecified, `druid.segmentCache.numThreadsToLoadSegmentsIntoPageCacheOnDownload` will be used. If both configs are unspecified, this feature is disabled. Preemptively loading segments into page cache helps in the sense that later when a segment is queried, it's already in page cache and only a minor page fault needs to be triggered instead of a more costly major page fault to make the query latency more consistent. Note that loading segment into page cache just does a blind loading of segment index files and will evict any existing segments from page cache at the discretion of operating system when the total segment size on local disk is larger than the page cache usable in the RAM, which roughly equals to total available RAM in the host - druid process memory including both heap and direct memory allocated - memory used by other non druid processes on the host, so it is the user's responsibility to ensure the host has enough RAM to host all the segments to avoid random evictions to fully leverage this feature.|`druid.segmentCache.numThreadsToLoadSegmentsIntoPageCacheOnDownload`|

Expand Down
18 changes: 3 additions & 15 deletions docs/design/zookeeper.md
Original file line number Diff line number Diff line change
Expand Up @@ -31,6 +31,7 @@ Apache Druid supports ZooKeeper versions 3.5.x and above.

:::info
Note: Starting with Apache Druid 0.22.0, support for ZooKeeper 3.4.x has been removed
Starting with Apache Druid 31.0.0, support for Zookeeper-based segment loading has been removed.
:::

## ZooKeeper Operations
Expand All @@ -39,9 +40,8 @@ The operations that happen over ZK are

1. [Coordinator](../design/coordinator.md) leader election
2. Segment "publishing" protocol from [Historical](../design/historical.md)
3. Segment load/drop protocol between [Coordinator](../design/coordinator.md) and [Historical](../design/historical.md)
4. [Overlord](../design/overlord.md) leader election
5. [Overlord](../design/overlord.md) and [MiddleManager](../design/middlemanager.md) task management
3. [Overlord](../design/overlord.md) leader election
4. [Overlord](../design/overlord.md) and [MiddleManager](../design/middlemanager.md) task management

## Coordinator Leader Election

Expand Down Expand Up @@ -74,15 +74,3 @@ ${druid.zk.paths.servedSegmentsPath}/${druid.host}/_segment_identifier_
```

Processes like the [Coordinator](../design/coordinator.md) and [Broker](../design/broker.md) can then watch these paths to see which processes are currently serving which segments.

## Segment load/drop protocol between Coordinator and Historical

The `loadQueuePath` is used for this.

When the [Coordinator](../design/coordinator.md) decides that a [Historical](../design/historical.md) process should load or drop a segment, it writes an ephemeral znode to

```
${druid.zk.paths.loadQueuePath}/_host_of_historical_process/_segment_identifier
```

This znode will contain a payload that indicates to the Historical process what it should do with the given segment. When the Historical process is done with the work, it will delete the znode in order to signify to the Coordinator that it is complete.
3 changes: 1 addition & 2 deletions docs/development/extensions-core/kubernetes.md
Original file line number Diff line number Diff line change
Expand Up @@ -31,11 +31,10 @@ Apache Druid Extension to enable using Kubernetes API Server for node discovery

To use this extension please make sure to [include](../../configuration/extensions.md#loading-extensions) `druid-kubernetes-extensions` in the extensions load list.

This extension works together with HTTP based segment and task management in Druid. Consequently, following configurations must be set on all Druid nodes.
This extension works together with HTTP-based segment and task management in Druid. Consequently, following configurations must be set on all Druid nodes.

`druid.zk.service.enabled=false`
`druid.serverview.type=http`
`druid.coordinator.loadqueuepeon.type=http`
`druid.indexer.runner.type=httpRemote`
`druid.discovery.type=k8s`

Expand Down

0 comments on commit 51104e8

Please sign in to comment.