SPARK-938 - Openstack Swift object storage support #1010
+270
−1
There are no files selected for viewing
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,110 @@ | ||
| yout: global | ||
| title: Accessing Openstack Swift storage from Spark | ||
| --- | ||
|
|
||
| # Accessing Openstack Swift storage from Spark | ||
|
|
||
| Spark's file interface allows it to process data in Openstack Swift using the same URI | ||
| formats that are supported for Hadoop. You can specify a path in Swift as input through a | ||
| URI of the form `swift://<container.service_provider>/path`. You will also need to set your | ||
| Swift security credentials, through `SparkContext.hadoopConfiguration`. | ||
|
|
||
| #Configuring Hadoop to use Openstack Swift | ||
| Openstack Swift driver was merged in Hadoop verion 2.3.0 ([Swift driver](https://issues.apache.org/jira/browse/HADOOP-8545)). Users that wish to use previous Hadoop versions will need to configure Swift driver manually. Current Swift driver | ||
| requieres Swift to use Keystone authentication method. There are recent efforts to support | ||
| also temp auth [Hadoop-10420](https://issues.apache.org/jira/browse/HADOOP-10420). | ||
| To configure Hadoop to work with Swift one need to modify core-sites.xml of Hadoop and | ||
| setup Swift FS. | ||
|
|
||
| <configuration> | ||
|
There was a problem hiding this comment. Is this needed? Can we just put this in core-site.xml under conf? (Basically removing the configuring Hadoop section) |
||
| <property> | ||
| <name>fs.swift.impl</name> | ||
| <value>org.apache.hadoop.fs.swift.snative.SwiftNativeFileSystem</value> | ||
| </property> | ||
| </configuration> | ||
|
|
||
| #Configuring Swift | ||
| Proxy server of Swift should include `list_endpoints` middleware. More information | ||
| available [here] (https://github.com/openstack/swift/blob/master/swift/common/middleware/list_endpoints.py) | ||
|
|
||
| #Configuring Spark | ||
| To use Swift driver, Spark need to be compiled with `hadoop-openstack-2.3.0.jar` | ||
| distributted with Hadoop 2.3.0. For the Maven builds, Spark's main pom.xml should include | ||
|
|
||
| <swift.version>2.3.0</swift.version> | ||
|
|
||
|
|
||
| <dependency> | ||
| <groupId>org.apache.hadoop</groupId> | ||
| <artifactId>hadoop-openstack</artifactId> | ||
| <version>${swift.version}</version> | ||
| </dependency> | ||
|
|
||
| in addition, pom.xml of the `core` and `yarn` projects should include | ||
|
|
||
| <dependency> | ||
| <groupId>org.apache.hadoop</groupId> | ||
| <artifactId>hadoop-openstack</artifactId> | ||
| </dependency> | ||
|
|
||
|
|
||
| Additional parameters has to be provided to the Swift driver. Swift driver will use those | ||
| parameters to perform authentication in Keystone prior accessing Swift. List of mandatory | ||
| parameters is : `fs.swift.service.<PROVIDER>.auth.url`, | ||
|
There was a problem hiding this comment. Might make sense to make a table or a bullet list for this, instead of just comma separate lists. |
||
| `fs.swift.service.<PROVIDER>.auth.endpoint.prefix`, `fs.swift.service.<PROVIDER>.tenant`, | ||
| `fs.swift.service.<PROVIDER>.username`, | ||
| `fs.swift.service.<PROVIDER>.password`, `fs.swift.service.<PROVIDER>.http.port`, | ||
| `fs.swift.service.<PROVIDER>.http.port`, `fs.swift.service.<PROVIDER>.public`, where | ||
| `PROVIDER` is any name. `fs.swift.service.<PROVIDER>.auth.url` should point to the Keystone | ||
| authentication URL. | ||
|
|
||
| Create core-sites.xml with the mandatory parameters and place it under /spark/conf | ||
| directory. For example: | ||
|
|
||
|
|
||
| <property> | ||
| <name>fs.swift.service.<PROVIDER>.auth.url</name> | ||
| <value>http://127.0.0.1:5000/v2.0/tokens</value> | ||
| </property> | ||
| <property> | ||
| <name>fs.swift.service.<PROVIDER>.auth.endpoint.prefix</name> | ||
| <value>endpoints</value> | ||
| </property> | ||
| <name>fs.swift.service.<PROVIDER>.http.port</name> | ||
| <value>8080</value> | ||
| </property> | ||
| <property> | ||
| <name>fs.swift.service.<PROVIDER>.region</name> | ||
| <value>RegionOne</value> | ||
| </property> | ||
| <property> | ||
| <name>fs.swift.service.<PROVIDER>.public</name> | ||
| <value>true</value> | ||
| </property> | ||
|
|
||
| We left with `fs.swift.service.<PROVIDER>.tenant`, `fs.swift.service.<PROVIDER>.username`, | ||
| `fs.swift.service.<PROVIDER>.password`. The best way to provide those parameters to | ||
|
There was a problem hiding this comment. This is still doable at runtime by setting the job conf, isn't it? |
||
| SparkContext in run time, which seems to be impossible yet. | ||
| Another approach is to adapt Swift driver to obtain those values from system environment | ||
| variables. For now we provide them via core-sites.xml. | ||
| Assume a tenant `test` with user `tester` was defined in Keystone, then the core-sites.xml | ||
| shoud include: | ||
|
|
||
| <property> | ||
| <name>fs.swift.service.<PROVIDER>.tenant</name> | ||
| <value>test</value> | ||
| </property> | ||
| <property> | ||
| <name>fs.swift.service.<PROVIDER>.username</name> | ||
| <value>tester</value> | ||
| </property> | ||
| <property> | ||
| <name>fs.swift.service.<PROVIDER>.password</name> | ||
| <value>testing</value> | ||
| </property> | ||
| # Usage | ||
| Assume there exists Swift container `logs` with an object `data.log`. To access `data.log` | ||
| from Spark the `swift://` scheme should be used. For example: | ||
|
|
||
| val sfdata = sc.textFile("swift://logs.<PROVIDER>/data.log") | ||
|
|
||
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
While it might make sense for this to eventually get into Spark, we need to look more carefully at the dependency that this brings. Since Spark runs different from Hadoop (Spark is really just a user level library), users can always include openstack support in their project dependencies (with the documentation you provide). For the time being, let's first update the documentation so it is obvious & clear & easy for users to add openstack support, and then we can discuss more about whether/when we should push the openstack dependency in Spark.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
There is now a change here that messes up spacing - @gilv can you remove this change?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Actually I'll just fix this up on merge. This is patrick by the way, logged into the QA account by accident!