diff --git a/hadoop-hdfs-project/hadoop-hdfs/src/site/markdown/OzoneCommandShell.md b/hadoop-hdfs-project/hadoop-hdfs/src/site/markdown/OzoneCommandShell.md new file mode 100644 index 0000000000..80bc46bea0 --- /dev/null +++ b/hadoop-hdfs-project/hadoop-hdfs/src/site/markdown/OzoneCommandShell.md @@ -0,0 +1,146 @@ + + +Ozone Command Shell +=================== + +Ozone command shell gives a command shell interface to work against ozone. +Please note that this document assumes that cluster is deployed +with simple authentication. + +The Ozone commands take the following format. + +* `hdfs oz --command_ http://hostname:port/volume/bucket/key -user + -root` + +The *--root* option is a command line short cut that allows *hdfs oz* +commands to be run as the user that started the cluster. This is useful to +indicate that you want the commands to be run as some admin user. The only +reason for this option is that it makes the life of a lazy developer more +easier. + +Ozone Volume Commands +-------------------- + +The volume commands allow users to create, delete and list the volumes in the +ozone cluster. + +### Create Volume + +Volumes can be created only by Admins. Here is an example of creating a volume. + +* `hdfs oz -createVolume http://localhost:9864/hive -user bilbo -quota +100TB -root` + +The above command creates a volume called `hive` owned by user `bilbo`. The +`--root` option allows the command to be executed as user `hdfs` which is an +admin in the cluster. + +### Update Volume + +Updates information like ownership and quota on an existing volume. + +* `hdfs oz -updateVolume http://localhost:9864/hive -quota 500TB -root` + +The above command changes the volume quota of hive from 100TB to 500TB. + +### Delete Volume +Deletes a Volume if it is empty. + +* `hdfs oz -deleteVolume http://localhost:9864/hive -root` + + +### Info Volume +Info volume command allows the owner or the administrator of the cluster to read meta-data about a specific volume. + +* `hdfs oz -infoVolume http://localhost:9864/hive -root` + +### List Volumes + +List volume command can be used by administrator to list volumes of any user. It can also be used by a user to list volumes owned by him. + +* `hdfs oz -listVolume http://localhost:9864/ -user bilbo -root` + +The above command lists all volumes owned by user bilbo. + +Ozone Bucket Commands +-------------------- + +Bucket commands follow a similar pattern as volume commands. However bucket commands are designed to be run by the owner of the volume. +Following examples assume that these commands are run by the owner of the volume or bucket. + + +### Create Bucket + +Create bucket call allows the owner of a volume to create a bucket. + +* `hdfs oz -createBucket http://localhost:9864/hive/january` + +This call creates a bucket called `january` in the volume called `hive`. If +the volume does not exist, then this call will fail. + + +### Update Bucket +Updates bucket meta-data, like ACLs. + +* `hdfs oz -updateBucket http://localhost:9864/hive/january -addAcl +user:spark:rw` + +### Delete Bucket +Deletes a bucket if it is empty. + +* `hdfs oz -deleteBucket http://localhost:9864/hive/january` + +### Info Bucket +Returns information about a given bucket. + +* `hdfs oz -infoBucket http://localhost:9864/hive/january` + +### List Buckets +List buckets on a given volume. + +* `hdfs oz -listtBucket http://localhost:9864/hive` + +Ozone Key Commands +------------------ + +Ozone key commands allows users to put, delete and get keys from ozone buckets. + +### Put Key +Creates or overwrites a key in ozone store, -file points to the file you want +to upload. + +* `hdfs oz -putKey http://localhost:9864/hive/january/processed.orc -file +processed.orc` + +### Get Key +Downloads a file from the ozone bucket. + +* `hdfs oz -getKey http://localhost:9864/hive/january/processed.orc -file + processed.orc.copy` + +### Delete Key +Deletes a key from the ozone store. + +* `hdfs oz -deleteKey http://localhost:9864/hive/january/processed.orc` + +### Info Key +Reads key metadata from the ozone store. + +* `hdfs oz -infoKey http://localhost:9864/hive/january/processed.orc` + +### List Keys +List all keys in an ozone bucket. + +* `hdfs oz -listKey http://localhost:9864/hive/january` diff --git a/hadoop-hdfs-project/hadoop-hdfs/src/site/markdown/OzoneGettingStarted.md b/hadoop-hdfs-project/hadoop-hdfs/src/site/markdown/OzoneGettingStarted.md new file mode 100644 index 0000000000..ed32aa75be --- /dev/null +++ b/hadoop-hdfs-project/hadoop-hdfs/src/site/markdown/OzoneGettingStarted.md @@ -0,0 +1,257 @@ + +Ozone - Object store for Hadoop +============================== + +Introduction +------------ +Ozone is an object store for Hadoop. It is a redundant, distributed object +store build by leveraging primitives present in HDFS. Ozone supports REST +API for accessing the store. + +Getting Started +--------------- +Ozone is a work in progress and currently lives in its own branch. To +use it, you have to build a package by yourself and deploy a cluster. + +### Building Ozone + +To build Ozone, please checkout the hadoop sources from github. Then +checkout the ozone branch, HDFS-7240 and build it. + +- `git checkout HDFS-7240` +- `mvn clean package -DskipTests=true -Dmaven.javadoc.skip=true -Pdist -Dtar -DskipShade` + +skipShade is just to make compilation faster and not really required. + +This will give you a tarball in your distribution directory. This is the +tarball that can be used for deploying your hadoop cluster. Here is an +example of the tarball that will be generated. + +* `~/apache/hadoop/hadoop-dist/target/hadoop-3.0.0-alpha4-SNAPSHOT.tar.gz` + +Please proceed to setup a hadoop cluster by creating the hdfs-site.xml and +other configuration files that are needed for your cluster. + +### Ozone Configuration + +Ozone relies on its own configuration file called `ozone-site.xml`. It is +just for convenience and ease of management -- you can add these settings +to `hdfs-site.xml`, if you don't want to keep ozone settings separate. +This document refers to `ozone-site.xml` so that ozone settings are in one +place and not mingled with HDFS settings. + + * _*ozone.enabled*_ This is the most important setting for ozone. + Currently, Ozone is an opt-in subsystem of HDFS. By default, Ozone is + disabled. Setting this flag to `true` enables ozone in the HDFS cluster. + Here is an example, + +``` + + ozone.enabled + True + +``` + * _*ozone.container.metadata.dirs*_ Ozone is designed with modern hardware + in mind. It tries to use SSDs effectively. So users can specify where the + datanode metadata must reside. Usually you pick your fastest disk (SSD if + you have them on your datanodes). Datanodes will write the container metadata + to these disks. This is a required setting, if this is missing datanodes will + fail to come up. Here is an example, + +``` + + ozone.container.metadata.dirs + /data/disk1/container/meta + +``` + +* _*ozone.scm.names*_ Ozone is build on top of container framework (See Ozone + Architecture TODO). Storage container manager(SCM) is a distributed block + service which is used by ozone and other storage services. + This property allows datanodes to discover where SCM is, so that + datanodes can send heartbeat to SCM. SCM is designed to be highly available + and datanodes assume there are multiple instances of SCM which form a highly + available ring. The HA feature of SCM is a work in progress. So we + configure ozone.scm.names to be a single machine. Here is an example, + +``` + + ozone.scm.names + scm.hadoop.apache.org + +``` + +* _*ozone.scm.datanode.id*_ Each datanode that speaks to SCM generates an ID +just like HDFS. This ID is stored is a location pointed by this setting. If +this setting is not valid, datanodes will fail to come up. Please note: +This path that is will created by datanodes to store the datanode ID. Here is an example, + +``` + + ozone.scm.datanode.id + /data/disk1/scm/meta/node/datanode.id + +``` + +* _*ozone.scm.block.client.address*_ Storage Container Manager(SCM) offers a + set of services that can be used to build a distributed storage system. One + of the services offered is the block services. KSM and HDFS would use this + service. This property describes where KSM can discover SCM's block service + endpoint. There is corresponding ports etc, but assuming that we are using + default ports, the server address is the only required field. Here is an + example, + +``` + + ozone.scm.block.client.address + scm.hadoop.apache.org + +``` + +* _*ozone.ksm.address*_ KSM server address. This is used by Ozonehandler and +Ozone File System. + +``` + + ozone.ksm.address + ksm.hadoop.apache.org + +``` + +Here is a quick summary of settings needed by Ozone. + +| Setting | Value | Comment | +|--------------------------------|------------------------------|------------------------------------------------------------------| +| ozone.enabled | True | This enables SCM and containers in HDFS cluster. | +| ozone.container.metadata.dirs | file path | The container metadata will be stored here in the datanode. | +| ozone.scm.names | SCM server name | Hostname:port or or IP:port address of SCM. | +| ozone.scm.datanode.id | file path | Data node ID is the location of datanode's ID file | +| ozone.scm.block.client.address | SCM server name | Used by services like KSM | +| ozone.ksm.address | KSM server name | Used by Ozone handler and Ozone file system. | + + Here is a working example of`ozone-site.xml`. + +``` + + + + + ozone.enabled + True + + + + ozone.container.metadata.dirs + /data/disk1/scm/meta + + + + + ozone.scm.names + scm.hadoop.apache.org + + + + ozone.scm.datanode.id + /data/disk1/scm/meta/node/datanode.id + + + + ozone.scm.block.client.address + scm.hadoop.apache.org + + + + ozone.ksm.address + ksm.hadoop.apache.org + + +``` + +### Starting Ozone + +Ozone is designed to run concurrently with HDFS. The simplest way to [start +HDFS](../hadoop-common/ClusterSetup.html) is to run `start-dfs.sh` from the +`$HADOOP/sbin/start-dfs.sh`. Once HDFS +is running, please verify it is fully functional by running some commands like + + - *./hdfs dfs -mkdir /usr* + - *./hdfs dfs -ls /* + + Once you are sure that HDFS is running, start Ozone. To start ozone, you + need to start SCM and KSM. Currently we assume that both KSM and SCM + is running on the same node, this will change in future. + + - `./hdfs --daemon start scm` + - `./hdfs --daemon start ksm` + +if you would like to start HDFS and Ozone together, you can do that by running + a single command. + - `$HADOOP/sbin/start-ozone.sh` + + This command will start HDFS and then start the ozone components. + + Once you have ozone running you can use these ozone [shell](./OzoneCommandShell.html) + commands to create a volume, bucket and keys. + +### Diagnosing issues + +Ozone tries not to pollute the existing HDFS streams of configuration and +logging. So ozone logs are by default configured to be written to a file +called `ozone.log`. This is controlled by the settings in `log4j.properties` +file in the hadoop configuration directory. + +Here is the log4j properties that are added by ozone. + + +``` + # + # Add a logger for ozone that is separate from the Datanode. + # + #log4j.debug=true + log4j.logger.org.apache.hadoop.ozone=DEBUG,OZONE,FILE + + # Do not log into datanode logs. Remove this line to have single log. + log4j.additivity.org.apache.hadoop.ozone=false + + # For development purposes, log both to console and log file. + log4j.appender.OZONE=org.apache.log4j.ConsoleAppender + log4j.appender.OZONE.Threshold=info + log4j.appender.OZONE.layout=org.apache.log4j.PatternLayout + log4j.appender.OZONE.layout.ConversionPattern=%d{ISO8601} [%t] %-5p \ + %X{component} %X{function} %X{resource} %X{user} %X{request} - %m%n + + # Real ozone logger that writes to ozone.log + log4j.appender.FILE=org.apache.log4j.DailyRollingFileAppender + log4j.appender.FILE.File=${hadoop.log.dir}/ozone.log + log4j.appender.FILE.Threshold=debug + log4j.appender.FILE.layout=org.apache.log4j.PatternLayout + log4j.appender.FILE.layout.ConversionPattern=%d{ISO8601} [%t] %-5p \ + (%F:%L) %X{function} %X{resource} %X{user} %X{request} - \ + %m%n +``` + +If you would like to have a single datanode log instead of ozone stuff +getting written to ozone.log, please remove this line or set this to true. + + ` log4j.additivity.org.apache.hadoop.ozone=false` + +On the SCM/KSM side, you will be able to see + + - `hadoop-hdfs-ksm-hostname.log` + - `hadoop-hdfs-scm-hostname.log` + +Please file any issues you see under [Object store in HDFS (HDFS-7240)](https://issues.apache.org/jira/browse/HDFS-7240) +as this is still a work in progress. diff --git a/hadoop-project/src/site/site.xml b/hadoop-project/src/site/site.xml index 1195862adc..92e460408e 100644 --- a/hadoop-project/src/site/site.xml +++ b/hadoop-project/src/site/site.xml @@ -109,6 +109,12 @@ + + + + + +