To close this thread, this turned out to be a user issue.
Any users who want to use impersonation can follow the suggested steps in the thread to do it. Users need to be careful about the client side classpath. The tez-api jars have META-INF properties that define the Hadoop security provider. If the client side classpath has issues that make this provider inaccessible then it can cause issues. Among other things, combining the tez jars in a shaded/uber jar can cause such issues. Thanks Subroto for bringing this up. This thread will be useful for anyone who wants to play with security using Tez/Hadoop. Bikas *From:* Bikas Saha [mailto:[email protected]] *Sent:* Wednesday, October 08, 2014 1:02 PM *To:* [email protected] *Subject:* RE: Tez with secured hadoop I have responded on the jira. Lets take the discussion there. *From:* Subroto Sanyal [mailto:[email protected]] *Sent:* Tuesday, October 07, 2014 10:44 PM *To:* [email protected] *Subject:* Re: Tez with secured hadoop hi, Please let me know how it is possible to get tokens and related stuffs in TEZ framework? On Tue, Oct 7, 2014 at 9:11 AM, Subroto Sanyal <[email protected]> wrote: hi Vinod, hi Bikas Thanks for your inputs. Though I am able to spawn the DAGAppMaster with proxy user (I can see the DAGAppMaster running with proxy user in resource manager UI) but, the call: TezClient.getAppMasterStatus fails with security issues. I have raised a ticket with my client code and more details: https://issues.apache.org/jira/browse/TEZ-1640 (UserGroupInformation.java:1551) - PriviledgedActionException as:qa (auth:SIMPLE) cause:java.io.IOException: org.apache.hadoop.security.AccessControlException: Client cannot authenticate via:[TOKEN, KERBEROS] Failed to retrieve AM Status via proxy com.google.protobuf.ServiceException: java.io.IOException: Failed on local exception: java.io.IOException: org.apache.hadoop.security.AccessControlException: Client cannot authenticate via:[TOKEN, KERBEROS]; Host Details : local host is: "ip-10-178-144-254/10.178.144.254"; destination host is: "ip-10-187-33-206":56660; at org.apache.hadoop.ipc.ProtobufRpcEngine$Invoker.invoke(ProtobufRpcEngine.java:216) at com.sun.proxy.$Proxy111.getAMStatus(Unknown Source) at org.apache.tez.client.TezClient.getAppMasterStatus(TezClient.java:522) On Tue, Oct 7, 2014 at 4:25 AM, Bikas Saha <[email protected]> wrote: Thanks Vinod. Subroto, looks like there are 2 issues here. One is running as use foo but submitting the job as user bar. Vinod’s code snippet is relevant to that (Please see my answer with createProxyUser earlier and Vinod’s exact code below). The real user is foo but the effective user is bar. So when the processes connects to YARN then it YARN will use the effective user bar as the job user instead of foo. Second is how to obtain credentials for bar while running as foo. That is covered by the tez-credentials-file-path solution mentioned earlier. This allows user foo to collect credentials from HDFS for user bar (foo is expected to be a trusted proxy user in HDFS to be allowed to do this. Like Oozie). Then the credentials are passed to the client via the file. Hope both these clarify your situation and move you forward. Bikas *From:* Vinod Kumar Vavilapalli [mailto:[email protected]] *Sent:* Monday, October 06, 2014 11:39 AM *To:* [email protected] *Subject:* Re: Tez with secured hadoop That is not the way to do impersonation. Please see the following: UserGroupInformation ugi = UserGroupInformation.createProxyUser("qa", UserGroupInformation.getLoginUser()); // <-- login user is subroto, the realUser. "qa" is the expected effective-user ugi.doAs(new PrivilegedExceptionAction<Void>() { public Void run() throws Exception { .. } } +Vinod Hortonworks Inc. http://hortonworks.com/ On Thu, Oct 2, 2014 at 6:22 AM, Subroto Sanyal <[email protected]> wrote: hi Bikas, My code snippet to create TezClient looks like (TEZ-0.5): new PrivilegedExceptionAction<TezClient>() { @Override public TezClient run() throws Exception { UserGroupInformation currentUser = UserGroupInformation.getCurrentUser(); LOG.info("Current User:" + currentUser); File tokenFile = new File(System.getProperty("java.io.tmpdir"), tezSessionName.replaceAll("[^a-zA-Z0-9]", "")); LOG.info("Token File:" + tokenFile.getAbsolutePath()); currentUser.getCredentials().writeTokenStorageFile(UriUtil.toPath(tokenFile.getAbsoluteFile()), conf); tezConf.set(TezConfiguration.TEZ_CREDENTIALS_PATH, tokenFile.getAbsolutePath()); return TezClient.create(tezSessionName, tezConf, createSession, localResourceMap, credentials); } } The logs generated from this piece of code during execution looks like: (TezClientFacade.java:142) - Current User:qa (auth:PROXY) via [email protected] (auth:KERBEROS) (TezClientFacade.java:144) - Token File:/home/subroto/tmp/testTezJob When this runs on cluster the job runs as "subroto" but, what I expect is to run it as "qa". Please let me know if there is something missing or wrong in the code. On Sat, Sep 13, 2014 at 3:33 AM, Bikas Saha <[email protected]> wrote: If by impersonation you mean what Oozie does where Oozie runs as Oozie but get delegation tokens for user FOO then you will need to follow the mechanism that Oozie uses. Oozie writes the delegation tokens into a file and put that file in a specific path which is picked up by the application (in this case TezClient) and the application loads credentials from that file. In case of Tez the location of the credentials file is the value of config "tez.credentials.path" Bikas *From:* Bikas Saha [mailto:[email protected]] *Sent:* Monday, September 01, 2014 5:34 PM *To:* [email protected] *Subject:* RE: Tez with secured hadoop They way this is supposed to work is the following in a secure cluster. 1) The user that is running TezClient/DAGClient needs to be Kerberos authenticated. This allows the process running DAGClient/TezClient to contact the RM and get tokens to communicate with the AM. 2) The TezClient/DAGClient uses the tokens obtained from the RM and populates it into the current user UGI (ie the use who is running TezClient/DAGClient). The RPC to the AM will try to authenticate the current user using the tokens just added to the current users UGI. In a non-secure environment, no tokens are needed. So I am guessing that your are running in a secure env. Given the above info, what is happening in your case. Whichever user the client is running under, it looks like it can authenticate to the RM to get the app report. So it should have gotten tokens to access the AM. Its not clear what you mean by user “subroto” being privileged and the real user not considered by Tez. It looks like you are running the client as user “subroto”. Who is “subroto” and who is the real user? Does this happen always or occasionally. There is a known race condition in YARN where the client gets tokens before the AM gets the key to validate the tokens. You can turn on debug logging and see the SASL negotiation logs to get more info on whats happening. You may add a debug log in getAMProxy() to verify that token were obtained from the RM and added to the UGI. It may help if you describe your scenario. What are you trying to achieve by impersonation and how are you trying to do that. We recently added ACLs in case that works for your scenario. *From:* Subroto Sanyal [mailto:[email protected]] *Sent:* Sunday, August 31, 2014 8:59 PM *To:* [email protected] *Subject:* Re: Tez with secured hadoop Hi Bikas, In the method: org.apache.tez.client.TezClientUtils.getAMProxy(Configuration, String, int, Token) a UGI is getting created with name of the current user. I think in this process it ignores all the security things and making the authentication mode as "SIMPLE". I have piece of code which tries to create a TezClient and it keeps throwing the exception: [anonymous] WARN [2014-08-28 03:37:50.181] [MrPlanRunnerV2] (UserGroupInformation.java:1551) - PriviledgedActionException as:subroto (auth:SIMPLE) cause:java.io.IOException: org.apache.hadoop.security.AccessControlException: Client cannot authenticate via:[TOKEN, KERBEROS] [anonymous] INFO [2014-08-28 03:37:50.182] [MrPlanRunnerV2] (TezClient.java:539) - Failed to retrieve AM Status via proxy com.google.protobuf.ServiceException: java.io.IOException: Failed on local exception: java.io.IOException: org.apache.hadoop.security.AccessControlException: Client cannot authenticate via:[TOKEN, KERBEROS]; Host Details : local host is: "domU-12-31-39-0F-74-32/10.193.119.192"; destination host is: "domU-12-31-39-0C-7D-37":59431; at org.apache.hadoop.ipc.ProtobufRpcEngine$Invoker.invoke(ProtobufRpcEngine.java:216) at com.sun.proxy.$Proxy111.getAMStatus(Unknown Source) at org.apache.tez.client.TezClient.getAppMasterStatus(TezClient.java:532) at org.apache.tez.client.TezClient.waitTillReady(TezClient.java:607) at subroto.tez.TezClusterSession$2.run(TezClusterSession.java:180) I m trying to achieve impersonation. Here user "subroto" is privileged user and the real user is not at all considered by the Tez Code. Request some suggestion on this. On Tue, Aug 19, 2014 at 11:18 PM, Bikas Saha <[email protected]> wrote: There is nothing special that you need to do if you are already running secure Map Reduce jobs. The client needs to run in a Kerberized authenticated context. After that if you are using the built-in library of inputs/outputs etc then they should be taking care of all the access credentials for you when using the 0.5 API. I If you are using 0.4 API to write your job then you may need to use additional APIs for passing credentials to the application. Look for credentials in https://github.com/apache/tez/blob/branch-0.4.0-incubating/tez-mapreduce-examples/src/main/java/org/apache/tez/mapreduce/examples/FilterLinesByWord.java and also *public* *synchronized* DAG *addURIsForCredentials(*Collection*<* URI*>* uris*)* The second method is a shortcut if you are using HDFS files for input. It obtains credentials for you from a collection of HDFS input URIs. Bikas *From:* Subroto Sanyal [mailto:[email protected]] *Sent:* Tuesday, August 19, 2014 3:30 AM *To:* [email protected] *Subject:* Tez with secured hadoop hi Tez works on secure hadoop cluster since tez-0.3. Is there any documentation available about configuring TezClient to make it work? -- Cheers, *Subroto Sanyal* CONFIDENTIALITY NOTICE NOTICE: This message is intended for the use of the individual or entity to which it is addressed and may contain information that is confidential, privileged and exempt from disclosure under applicable law. If the reader of this message is not the intended recipient, you are hereby notified that any printing, copying, dissemination, distribution, disclosure or forwarding of this communication is strictly prohibited. If you have received this communication in error, please contact the sender immediately and delete it from your system. Thank You. -- Cheers, *Subroto Sanyal* CONFIDENTIALITY NOTICE NOTICE: This message is intended for the use of the individual or entity to which it is addressed and may contain information that is confidential, privileged and exempt from disclosure under applicable law. If the reader of this message is not the intended recipient, you are hereby notified that any printing, copying, dissemination, distribution, disclosure or forwarding of this communication is strictly prohibited. If you have received this communication in error, please contact the sender immediately and delete it from your system. Thank You. CONFIDENTIALITY NOTICE NOTICE: This message is intended for the use of the individual or entity to which it is addressed and may contain information that is confidential, privileged and exempt from disclosure under applicable law. If the reader of this message is not the intended recipient, you are hereby notified that any printing, copying, dissemination, distribution, disclosure or forwarding of this communication is strictly prohibited. If you have received this communication in error, please contact the sender immediately and delete it from your system. Thank You. -- Cheers, *Subroto Sanyal* CONFIDENTIALITY NOTICE NOTICE: This message is intended for the use of the individual or entity to which it is addressed and may contain information that is confidential, privileged and exempt from disclosure under applicable law. If the reader of this message is not the intended recipient, you are hereby notified that any printing, copying, dissemination, distribution, disclosure or forwarding of this communication is strictly prohibited. If you have received this communication in error, please contact the sender immediately and delete it from your system. Thank You. CONFIDENTIALITY NOTICE NOTICE: This message is intended for the use of the individual or entity to which it is addressed and may contain information that is confidential, privileged and exempt from disclosure under applicable law. If the reader of this message is not the intended recipient, you are hereby notified that any printing, copying, dissemination, distribution, disclosure or forwarding of this communication is strictly prohibited. If you have received this communication in error, please contact the sender immediately and delete it from your system. Thank You. -- Cheers, *Subroto Sanyal* -- Cheers, *Subroto Sanyal* -- CONFIDENTIALITY NOTICE NOTICE: This message is intended for the use of the individual or entity to which it is addressed and may contain information that is confidential, privileged and exempt from disclosure under applicable law. If the reader of this message is not the intended recipient, you are hereby notified that any printing, copying, dissemination, distribution, disclosure or forwarding of this communication is strictly prohibited. If you have received this communication in error, please contact the sender immediately and delete it from your system. Thank You.
