Hello Guillaume,
What about extending org.apache.hadoop.hbase.HBaseClusterTestCase,
instantiating it and invoking its setUp()? I know it isn't the most
beautiful way to start HBase from java, but it works pretty well here.
Something along the lines of:
public class StandAloneServer extends HBaseClusterTestCase {
public StandAloneServer(Properties props) {
super();
// iterate over props map invoking: conf.set(propKey, propValue);
}
public static void main(String[] args) {
Properties props;
// load props from args[0], for example
StandAloneServer sas = new StandAloneServer(props);
// add close hook
Runtime.getRuntime().addShutdownHook(new Thread() {
public void run() {
sas.tearDown();
}
});
sas.setUp();
}
}
2009/10/13 Guillaume Dufrêne <[email protected]>
> stack wrote:
>
>> A few comments interspersed in the below:
>>
>> 2009/10/13 Guillaume Dufrêne <[email protected]>
>>
>>
>>
>>
>>> I wrote a class inspired from the Mini*Server located in Hbase Tests.
>>>
>>> (I use Hbase 0.20)
>>>
>>> Here my class :
>>>
>>> public class StandaloneServer {
>>>
>>> private static Log log = LogFactory.getLog(StandaloneServer.class);
>>>
>>> private MiniZooKeeperCluster zookeeper;
>>> private MiniHBaseCluster hbase;
>>> private MiniDFSCluster dfs;
>>>
>>> private HBaseConfiguration conf;
>>>
>>> public void startup() throws Exception {
>>> conf = new HBaseConfiguration();
>>>
>>> log.info("Start DFS ...");
>>> dfs = new MiniDFSCluster(conf, 2, true, (String[]) null);
>>> // set file system to the mini dfs just started up
>>> FileSystem fs = dfs.getFileSystem();
>>> conf.set("fs.default.name", fs.getUri().toString());
>>>
>>>
>>>
>>
>> You are using 0.20? FYI, the above name is deprecated in TRUNK. HBase
>> TRUNK has been changed to use new name.
>>
>>
>>
>>
> Yep, I use 0.20 release, I didn't build hbase myself from the trunk.
> It seems that "fs.default.name" is used in the 0.20 release as shown in
> unit tests.
>
>
> Path parentdir = fs.getHomeDirectory();
>>> conf.set(HConstants.HBASE_DIR, parentdir.toString());
>>> fs.mkdirs(parentdir);
>>> FSUtils.setVersion(fs, parentdir);
>>>
>>> log.info("Start Zookeeper ...");
>>> zookeeper = new MiniZooKeeperCluster();
>>> int clientPort = zookeeper.startup(new File(parentdir.toString(),
>>> "zookeeper"));
>>> conf.set("hbase.zookeeper.property.clientPort", Integer
>>> .toString(clientPort));
>>> // Disable regionserver info.
>>> conf.set("hbase.regionserver.info.port", "-1");
>>> conf.set("hbase.master.info.port", "-1");
>>>
>>> log.info("Start HBase ...");
>>> hbase = new MiniHBaseCluster(conf, 1);
>>>
>>> log.info("Check ROOT metatable connexion ...");
>>> new HTable(conf, HConstants.META_TABLE_NAME);
>>>
>>> }
>>>
>>> public void shutdown() {
>>> hbase.shutdown();
>>> if (dfs != null) {
>>> try {
>>> FileSystem fs = dfs.getFileSystem();
>>> if (fs != null)
>>> fs.close();
>>> } catch (IOException e) {
>>> System.err.println("error closing file system: " + e);
>>> }
>>> try {
>>> dfs.shutdown();
>>> } catch (Exception e) { /* ignore */
>>> }
>>> }
>>> }
>>>
>>>
>>>
>>
>> The shutdown thread stuff is different in TRUNK.
>>
>>
>
> Thanks for that information, I will update my code when I migrate hbase
> version.
>
>
> public static void main(String[] args) throws Exception {
>>> final StandaloneServer server = new StandaloneServer();
>>> server.startup();
>>>
>>> // add close hook
>>> Runtime.getRuntime().addShutdownHook(new Thread() {
>>> public void run() {
>>> server.shutdown();
>>> }
>>> });
>>>
>>> }
>>> }
>>>
>>> I first wrote a main operation to easily test the behaviour ...
>>> The idea is to start the server when my unit test is set up.
>>>
>>> I got the following error :
>>> 09/10/13 14:12:13 INFO hbase.StandaloneServer: Check ROOT metatable
>>> connexion ...
>>> Exception in thread "main"
>>> org.apache.hadoop.hbase.client.NoServerForRegionException: No server
>>> address
>>> listed in -ROOT- for region .META.,,1
>>> at
>>>
>>> [oahh.client]HConnectionManager$TableServers.locateRegionInMeta(HConnectionManager.java:660)
>>>
>>> at
>>>
>>> [oahh.client].HConnectionManager$TableServers.locateRegion(HConnectionManager.java:582)
>>>
>>> at
>>>
>>> [oahh.client].client.HConnectionManager$TableServers.locateRegion(HConnectionManager.java:549)
>>>
>>> at [oahh.client].HTable.<init>(HTable.java:125)
>>> at (...) StandaloneServer.startup(StandaloneServer.java:53)
>>> at (...) StandaloneServer.main(StandaloneServer.java:76)
>>>
>>>
>>>
>>
>>
>> This happens on start up until root is deployed... you'll see one or two
>> in
>> the logs.
>>
>>
> The StackTrace is triggered from my last line of code that check access to
> META table.
> The MiniHBaseCluster is supposed to wait for region server to start, isn't
> it ?
> Does the "root" you pointed out be the HBase master ? (I suppose so)
> What should I do to ensure this root deployment ?
> do a try/catch and retry N times with a delay ?
>
> And some errors about "NO JSP Support" (but I think that I can ignore
>>> them,
>>> no ?)
>>>
>>>
>>>
>> You've turned off the info servers -- the web uis -- so should be fine.
>>
>>
>>
>>
>>
>>> So my question are :
>>> - Is it the good way to start a test environment ?
>>>
>>>
>>>
>>
>> Yeah, you seem to have figured out our ugly test setup. You could also
>> have
>> run hbase with all defaults? Then it runs a master and regionserver in
>> same
>> in JVM writing the local fileystem.
>>
>>
> You mean that I can run "LocalHBaseCluster" with a new HBaseConfiguration()
> without starting FS ?
>
> If Someone has a java code sample that start a hbase, let's share ;-)
>
>> [...]
>>>
>>
> Thanks.
> --
> Guillaume.
>
>
>
>
>
>
> Ce message et toutes les pièces jointes sont établis à l'attention
> exclusive de leurs destinataires et sont confidentiels. Si vous recevez ce
> message par erreur, merci de le détruire et d'en avertir immédiatement
> l'expéditeur. L'internet ne permettant pas d'assurer l'intégrité de ce
> message, le contenu de ce message ne représente en aucun cas un engagement
> de la part de Leroy Merlin.
>
>
--
Guilherme
msn: [email protected]
homepage: http://sites.google.com/site/germoglio/
