I do not have access to any CLI on any of the Hadoop nodes, but I have access to the cluster via Hue and Jupyter. The engineering team has also configured the Hadoop UI that shows New, Running, Submitted, Finish, etc. applications. However, it appears all spark jobs have a generic name, for instance, something like this:
HIVE-f23fa1a1-4444-4ab2-1c44-12345a123456
or similar and when I click on the application_id, I get a Failed to read the attempts of the application error. (even for my own jobs). Similarly, spark jobs, which you can normally name using setAppName, are all named generic "Spark-something" because the spark context is already initialized upon bringing up Jupyter on an edge node (i.e. I can't establish a name because one already exists).
Is there a way for a unprivileged Hadoop user to see into what job is actually running (i.e. the Hive query or the Spark / Hadoop command ), without having some sort of CLI privilege?
I have tried using a few urls that I suspect have job information in them, for instance:
http://cluster_master:<portnum>/history/application_1234123412341234_12345/jobs/ or
http://cluster_master:<portnum>/jobs/application_1234123412341234_12345/
but neither attempt returns any details about the job itself (even things I named myself within the hive / spark context using setAppName.
Please let me know if there's a better way to ask this question. I am relatively new to Hadoop/Spark. All the reference docs and SO answers I've found assume CLI or privileged access and I can't find any documentation in either Spark or Hadoop that applies to this problem.
Related
My question is pretty trivial but didnt find anyone actually asking it.
We have a ambari cluster with spark storm hbase and hdfs(among other things).
I dont understand how a user that want to use that cluster use it.
for example, a user want to copy a file to hdfs, run a spark-shell or create new table in hbase shell.
should he get a local account on the server that run the cooresponded service? shouldn't he use a 3rd party machine(his own laptop for example)?
If so ,how one should use hadoop fs, there is no way to specify the server ip like spark-shell has.
what is the normal/right/expected way to run all these tasks from a user prespective.
Thanks.
The expected way to run the described tasks from the command line is as follows.
First, gain access to the command line of a server that has the required clients installed for the services you want to use, e.g. HDFS, Spark, HBase et cetera.
During the process of provisioning a cluster via Ambari, it is possible to define one or more servers where the clients will be installed.
Here you can see an example of an Ambari provisioning process step. I decided to install the clients on all servers.
Afterwards, one way to figure out which servers have the required clients installed is to check your hosts views in Ambari. Here you can find an example of an Ambari hosts view: check the green rectangle to see the installed clients.
Once you have installed the clients on one or more servers, these servers will be able to utilize the services of your cluster via the command line.
Just to be clear, the utilization of a service by a client is location-independent from the server where the service is actually running.
Second, make sure that you are compliant with the security mechanisms of your cluster. In relation to HDFS, this could influence which users you are allowed to use and which directories you can access by using them. If you do not use security mechanisms like e.g. Kerberos, Ranger and so on, you should be able to directly run your stated tasks from the command line.
Third, execute your tasks via command line.
Here is a short example of how to access HDFS without considering security mechanisms:
ssh user#hostxyz # Connect to the server that has the required HDFS client installed
hdfs dfs -ls /tmp # Command to list the contents of the HDFS tmp directory
Take a look on Ambari views, especially on Files view that allows browsing HDFS
What I would like to do is make workflow and job metadata such as start date, end date and status available in a hive table to be consumed by a BI tool for visualization purposes. I would like to be able to monitor for example if a certain workflow fails on certain hours, success rate, ...
For this purpose I need access to the same data Hue is able to show in the job browser and Oozie dashboard. What I am looking for specifically for workflows for example is the name, submitter, status, start and end time. The reason that I want this is that in my opinion this tool lacks a general overview and good search.
The idea is that once I locate this data I will directly -or trough some processing steps- load it into Hive.
Questions that I would like to see answered:
Is this data stored in HDFS or is it scattered in local data nodes?
If it is stored in HDFS. Where can I find it? If it is stored in local data nodes, how does Hue find and show this?
Assuming I can access the data. In what format would I expect this data. Is this stored in general log files or can I expect somewhat structured data?
I am using CDH 5.8
If jobs are submitted through other ways than Oozie , my approach won't be helpful.
We have collected all the logs from the oozie server through the Oozie Java API and iterated over the coordinator information to get the required info.
You need to think, what kind of information you need to retrieve.
If you have all jobs submitted through Bundle then come from bundle to coordinator then to workflow to find out the info.
If you want to get all the coordinator info then simply call the api with the number of coordinator to bring and fetch required info.
And then we have loaded the fetched result into a hive table and there one can filter results for failed or time out coordinators & various other parameters.
You can start looking into the example given from Oozie site:-
https://oozie.apache.org/docs/3.2.0-incubating/DG_Examples.html#Java_API_Example]
If you want to track the status of your jobs scheduled in oozie, you should use oozie RESTful API or JavaAPI. I didn't work with Hue version for operation Oozie, but I guess it still uses rest api behind the scene. It provides you with all necessary information and you can create some service which would consume this data and push it into Hive table.
Another option is to access Oozie database. As you probably know Oozie keeps all the data about the scheduled jobs within some RDBMS like MqSql or Postgres. You can consume this information through some JDBC connector. An interesting way would actually be to try to link this information directly into Hive as a set of external tables though JDBCStorageHandler. Not sure if it work, but it worth to try.
We are using a kerborized CDH cluster. While adding a user to the cluster, we used to add the user only to the gateway/edge nodes as in any hadoop distro cluster. But with the newly added userIDs, we are not able to execute map-reduce/yarn jobs and throwing "user not found" exception.
When I researched through this, I came across a link https://hadoop.apache.org/docs/stable/hadoop-yarn/hadoop-yarn-site/SecureContainer.html , which says to execute the yarn jobs in the secured cluster, we might need to have the corresponding user in all the nodes as the secure containers execute under the credentials of the job user.
So we added the corresponding userID to all the nodes and the jobs are getting executed.
If this is the case and if the cluster has around 100+ nodes, user provisioning for each userID would became a tedious job.
Can anyone please suggest any other effective way, if you came across the same scenario in your project implementation?
There are several approaches ordered by difficulty (from simple to painful).
One is to have a job-runner user that everyone uses to run jobs.
Another one is to use a configuration management tool to sync /etc/passwd and /etc/group (chef, puppet) on your cluster at regular intervals (1 hr - 1 day) or use a cron job to do this.
Otherwise you can buy or use open source Linux/UNIX user mapping services like Centrify (commercial), VAS (commercial), FreeIPA (free) or SSSD (free).
If you have an Active Directory server or LDAP server use the Hadoop LDAP user mappings.
References:
https://community.hortonworks.com/questions/57394/what-are-the-best-practises-for-unix-user-mapping.html
https://www.cloudera.com/documentation/enterprise/5-9-x/topics/cm_sg_ldap_grp_mappings.html
I trying to set up a hive environment on my google compute engine hadoop clusters which was deployed from one click deployment.
When I try to switch to hdfs user(su hdfs), I get below error message.
No passwd entry for user 'hdfs'
The "one-click deployment" is an older sample which perhaps showcases installation from shell scripts and tarballs, but isn't intended for use as a supported Hadoop service, and doesn't set up typical Hadoop installation configurations like an hdfs user or adding commands to /usr/bin.
If you want a more Hadoop (and Pig+Hive+Spark) specialized service, you may want to consider using Google Cloud Dataproc, which is Google's managed Hadoop solution. You can create clusters from the cloud console UI in Dataproc just like click-to-deploy, and you'll get a more fully installed Hadoop/Hive environment, including a per-cluster persistent MySQL-based Hive metastore which is shared with SparkSQL to make it easy to play with Spark without modifying your Hive environment if you so choose.
I've just created a google cloud dataproc cluster. A few basic things are not working for me:
I'm trying to run the hive console from the master node but it fails to load with any user other than root (it looks like there's a lock, the console is just stuck).
But even when using root, I see some odd behaviour:
"show tables;" shows a table named "input"
querying the table raises an exception that this table not found.
It is not clear which user is creating the tables through the web ui. I create a job, execute it, but then don't see the results through the console.
Couldn't find any good documentation on that - does anybody have an idea on this?
Running the hive command at present is somewhat broken due to the default metastore configuration.
I recommend you use the beeline client instead, which talks to the same Hive Server 2 as Dataproc Hive Jobs. You can use it via ssh by running beeline -u jdbc:hive2://localhost:10000 on the master.
YARN applications are submitted by the Hive Server 2 as the user "nobody", you can specify a different user by passing the -n flag to beeline, but it shouldn't matter with default permissions.
This thread is a bit old but when some one search Google Cloud Platform and Hive this result is coming. So I'm adding some info which may be useful.
Currently, in order to submit job to Google dataproc, I think - like all other products - there are 3 options:
from UI
from console using command line like:
gcloud dataproc jobs submit hive --cluster=CLUSTER (--execute=QUERY, -e QUERY | --file=FILE, -f FILE) [--async] [--bucket=BUCKET] [--continue-on-failure] [--jars=[JAR,…]] [--labels=[KEY=VALUE,…]] [--params=[PARAM=VALUE,…]] [--properties=[PROPERTY=VALUE,…]] [GLOBAL-FLAG …]
REST API call like: https://cloud.google.com/dataproc/docs/reference/rest/v1/projects.regions.jobs/submit
Hope this will be useful to someone.