Ask Your Question

dmitrymex's profile - activity

2014-12-01 16:46:07 -0500 answered a question "No module named config" installing Sahara Juno on Ubuntu 14.04

I think that is the issue with namespaces: your python does not see oslo.config, while it is installed. Try reinstalling it, like

pip uninstall oslo.config
pip install oslo.config

If it does not help, probably just plain reinstall from scratch will help.

2014-10-11 03:08:14 -0500 commented question heat authentication failure in devstack

I assume that you show heat-api log above ^^. Do you see anything in heat-engine log? It is my understanding that the AuthorizationFailure happened on heat-engine side and probably there are more details in the log.

2014-08-21 11:01:29 -0500 received badge  Nice Answer (source)
2014-08-13 08:37:46 -0500 received badge  Nice Answer (source)
2014-08-13 05:14:19 -0500 answered a question Sahara: Error database when reboot server

I think that your /tmp is cleaned during the reboot. Try moving sahara.db to a different directory.

2014-08-13 05:09:37 -0500 commented question Enable Spark Plugin

I know that do supports spark plugin, but I don't know more. Consider creating another question to bring more attention. Or maybe it would be easier and faster to ask via IRC in #openstack-sahara on freenode.

2014-08-12 11:06:13 -0500 commented question Enable Spark Plugin

I recommend not to perform 'pip install' manually. tox uses its own virtual envs (one could be found in .tox/venv) and automatically installs all the requirements there. I.e. steps #3 and #4 in your section 'Edit 1' could be safely omitted.

2014-08-12 05:43:59 -0500 answered a question Enable Spark Plugin

You face two problems:

Spark plugin is not available in Icehouse, it will be first released in Juno. Right now it can be used from the master branch.

The sahara-all binary appeared only in Juno. In Icehouse the same binary is called sahara-api and you should invoke it. I think that you still can invoke sahara-all only because you have .pyc files from your previous attempts with master.

2014-08-12 04:06:46 -0500 commented answer Sahara: Can't login to node

One more possible reason: you didn't configure security group properly. You say that you can ssh to instances from the host where Sahara is installed and using the IP addresses Sahara uses to access instances. For example, in the logs you provided these are and

2014-07-15 14:55:52 -0500 commented answer Sahara: Master-instance: Service hadoop-* not running

Did you try all the commands together:

sudo su hadoop
cd ~
hadoop jar /usr/share/hadoop/hadoop-examples-1.2.1.jar pi 10 100

I expect it to work because I think your problem was that you were running 'hadoop' command in catalog where you had no write permission. Also I think it is important to run it under 'hadoop' user.

2014-07-15 11:31:47 -0500 answered a question Sahara: Master-instance: Service hadoop-* not running

Ok, you are doing alright. You just need to specify full path to the hadoop-examples.jar file, like that:

hadoop jar /usr/share/hadoop/<...>/hadoop-examples-1.2.1.jar pi 10 100

Don't forget to replace '<...>' in command I've proposed with the actual path, so that /usr/share/hadoop/<...>/hadoop-examples-1.2.1.jar a full path for hadoop-examples.jar.

2014-07-14 16:34:32 -0500 commented question Sahara: Master-instance: Service hadoop-* not running

The info in the question is not full, but let me take a wild guess: if you execute sudo su hadoop; cd ~

and run Hadoop job after that, does the job work?

If that does not help, let me ask several questions: 1. Which plugin do you use: Vanilla or HDP? 2. Which Hadoop version do you install? 3. Sahara uses prepared images. Which image did you downloaded and used?

2014-07-04 04:36:30 -0500 received badge  Citizen Patrol
2014-06-26 00:54:05 -0500 received badge  Nice Answer (source)
2014-06-22 11:59:29 -0500 commented question when will the python api be released for openstack

The wiki page also lists python-openstackclient. This is the effort to unite all the other clients in one single client. I am not sure if it is finished. But you definitely can use 'personal' clients like python-keystoneclient, python-novaclient, etc.

2014-06-22 11:56:58 -0500 commented question when will the python api be released for openstack

Don't trust the wiki page, it is outdated :-). The clients which are marked as 'in development' there (cinder, heat) are already finished and ready for use. _Every_ official OpenStack project has a client library with name python-<something>client, for example - python-keystoneclient. See the link to the article I've posted above for the examples.

2014-06-22 10:59:38 -0500 received badge  Commentator
2014-06-22 10:59:38 -0500 commented question when will the python api be released for openstack

It is already there for a long time. For example see the following article: (it is outdated, so replace 'quantum' with 'neutron').

If it is not what you are looking for, please clarify your question.

2014-06-20 01:53:22 -0500 answered a question Sahara Configuration file

In fact 'host' is called 'bind_host' in many other projects. For example, you have two network IP address on your machine: and

Then if you set 'host=', Sahara process will be accessible only by that IP. I.e. if you do curl<tenant_id>/clusters you will get some response. But if you do curl<tenant_id>/clusters connection will be refused.

Most probably you want Sahara to be accessible on all configured IPs. To do that, just leave that field empty.

2014-06-18 04:24:58 -0500 commented answer Sahara installation on virtual instance

Yes, you can do that. It will work fine, just make sure VM can access OpenStack services (nova, glance, etc.) REST API.

2014-06-18 04:00:19 -0500 commented answer Sahara installation on virtual instance

No, currently Sahara is not integrated with Ceilometer.

2014-06-17 05:45:59 -0500 commented answer Sahara installation on virtual instance

When spawning VMs, Sahara injects SSH keypair on them which will be later used to provision Hadoop onto VMs. You should make sure that Sahara can access the port 22 of the VMs it spawned.

2014-06-17 05:43:09 -0500 commented answer Sahara installation on virtual instance

You actually just tell Sahara: run one 'master' node (JobTracker + NameNode) + N 'worker' nodes (TaskTracker + DataNode). Sahara then itself starts up N+1 VMs using the Nova REST APIs for the cluster.

As for VM characteristics, you set this as a flavor on Node Group level. Say, you can use m1.medium (2 cpu, 4 Gig RAM) for master node group (1 VM) and m1.small (1 cpu, 2 Gig RAM) for worker node group (N VMs)

2014-06-16 10:05:00 -0500 commented answer Sahara installation on virtual instance

After you install Sahara on Controller, you should be able to provision virtual instances with Hadoop cluster installed on them. Watch the following video on youtube to see how you can use Sahara after it is installed.

2014-06-16 09:54:39 -0500 answered a question Sahara with Ceph

Sahara can work with Ceph via Ceph Object Gateway. We use such setup in Fuel and it works well there without any modification of Sahara. Don't forget to register the Gateway in Keystone as object storage endpoint.

2014-06-10 22:28:29 -0500 received badge  Nice Answer (source)
2014-06-09 11:13:44 -0500 received badge  Nice Answer (source)
2014-06-09 11:13:05 -0500 received badge  Nice Answer (source)
2014-06-05 08:50:22 -0500 commented answer Devstack sahara installation

Maybe I misunderstood your initial question. The Sahara service itself will be run together with other OpenStack services like Nova, Glance, etc on the same host. But as for Hadoop, Sahara will provision it on VMs.

Does that answer your question?

2014-06-05 08:17:56 -0500 answered a question Devstack sahara installation

Sahara will be installed as part of the OpenStack. You can check that by running 'ps aux | grep sahara' after you've started the DevStack. The output will show sahara-api process running on your DevStack host.

2014-06-04 05:01:40 -0500 received badge  Editor (source)
2014-06-04 03:38:17 -0500 answered a question Devstack Horizon sahara failed

Thy setting Sahara service name all lowercase in localrc:


In my case it worked and when I tried 'Sahara', I didn't even get Sahara menu in UI

2014-06-02 05:04:51 -0500 answered a question How to login to Sahara VM via broswer console ?

I think when you tell "I can able to login from unix prompt using cloud-user and it did not ask for password via unix terminal" you mean you accessed the VM using SSH and you supplied the proper SSL keypair, right? In that case the VM authenticates you by the keypair you've provided.

In your browser you log in via VNC console. Here you can not supply SSL keypair, you can only use password to log in. I am not sure, but I think that all images provided by us do not have password set for any user. So I see two ways for you to proceed:

  • Log into VM using SSH and set password for cloud user using 'passwd' command. You will be able later to log in via browser as cloud-user using password you have set.
  • Build your own image using our diskimage-builder elements and here add your own element which will set password for the user. See for the details

Hope it is not too late :-(

2014-05-05 02:10:47 -0500 received badge  Teacher (source)
2014-04-16 04:16:03 -0500 answered a question openstack-sahara no avaliable

Sahara will be released only tomorrow along with the OpenStack Icehouse release. After the release, Installation instructions should start working, _maybe_ except RDO. As far as I know RDO packages are published a little bit later after the release date.

Sahara is a new name of Savanna project. The project was renamed to avoid trademark issues a month ago. So right now you can find only Savanna packages released. If you are installing OpenStack Havana, you may follow these instructions: They correspond to Savanna 0.3 (a previous release of Savanna/Sahara) which works with OpenStack Havana.

2014-04-03 05:19:06 -0500 answered a question keystoneclient.middleware.auth_token [-] Configuring auth_uri to point to the public identity endpoint is required

Discussing the same question in Ubuntu Forums:

2014-04-02 05:51:26 -0500 commented answer Glance gateway timeout error while uploading Images

I am not an expert in Glance, so I can't suggest where to find that timeout setting.

2014-04-02 05:46:33 -0500 answered a question Internel Server Error when installing Savanna

The same question is answered on Ubuntu forums:

My overall suggestion would be to use DevStack instead of CloudGear. The later it not maintained for a year and installs Grizzly which is outdated. DevStack on the contrary always installs the latest version.

2014-03-28 08:22:53 -0500 received badge  Supporter (source)
2014-03-27 05:26:33 -0500 answered a question Glance gateway timeout error while uploading Images

Hmm, I am not an expert in Glance, but I can suggest a possible workaround: try uploading from one of controller nodes via CLI:

glance image-create --name=test_image --disk-format=qcow2 --container-format=bare < your_image.qcow2

2014-03-26 03:01:40 -0500 answered a question Savanna/Sahara: How to assign a floating_ip_pool on node group?

Here is what you should do:

  1. Set the networking parameters to the following values in savanna.conf:
  2. Remove 'floating_ip_pool', Savanna does not have such parameter in config file.
  3. Make sure that the parameters above are not redefined somewhere else in the config file and restart Savanna.
  4. In /etc/openstack-dashboard/local_settings set.
  6. Restart Horizon.

After that when you create a new node group template in the UI, you should see a new 'Floating IP pool' parameter.

2014-03-26 02:30:23 -0500 commented question Glance gateway timeout error while uploading Images

How to do you upload the image? I know to ways:

  • While creating an image, provide an URL to the image location.
  • Download image manually and upload it as a file via glance CLI utility.
2014-03-26 02:17:44 -0500 received badge  Enthusiast
2014-03-25 16:18:05 -0500 commented question "qcow2" image with Openstack in Multi Node System

As far as I can see, you correctly import image into glance.

Did you try to redownload the image? I've seen images which didn't boot because they were corrupted. Though the errors were never that strange. In my case instances simply hung with message 'Booting from hard drive'

2014-03-14 15:03:27 -0500 answered a question how to upload image to glance(grizzly) from local disk through rest api in python.

Glance, like all other OpenStack services has a python client:

Using clients from python is generally much easier than sending http requests via httplib2 or requests libs. You can find an example on how to use glance and other python clients in the following article:

2014-03-13 04:33:20 -0500 answered a question "savanna-db-manage" error "invalid choice: upgrade"

Savanna version 0.3 corresponds to the OpenStack Havanna, so I assume you are using it.

Note that the guide you've posted corresponds to the master branch. You can browse docs for non-master at readthedocs. Here is the link to 0.3 documentation:

Here you will see that you don't need to run savanna-db-manage to start Savanna. In 0.3 the database is lazily created on startup if it doesn't exist.