ambari-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From cs user <acldstk...@gmail.com>
Subject Re: Exporting cluster configuration blueprint - recreation
Date Wed, 25 Nov 2015 09:29:30 GMT
Hi Naga,

Thanks for getting back to me. So the reason this was failing was due to
the fact I had stripped much of the content out of the blueprint (by
setting each content field to be "content" : "  ") , as I was having
trouble passing this as data over ssh (via the chef knife command).

Effectively this was setting the content, and so the config, to be
empty...... :-)

I've realized that if I just remove these content directives, ambari will
set these to the default values :-) I'm also base64 encoding the data and
then decoding it in the chef recipe before performing the curl.

Cheers!






On Tue, Nov 24, 2015 at 4:08 PM, Naga Vijay <nagahive@gmail.com> wrote:

> Hi cs user,
>
> I haven't faced this issue, as I use a custom AMI in which JAVA_HOME is
> set before hand, as part of the environment setup, so when the EC2 instance
> comes up, it is all set.
>
> Thanks
> Naga
>
>
> On Tue, Nov 24, 2015 at 7:56 AM, cs user <acldstkusr@gmail.com> wrote:
>
>> Hi Naga,
>>
>> The host to host_group mappings are defined in the cluster setup call.
>>
>> I'm having a problem however which I never encountered when I performed a
>> manual installation.
>>
>> I'm running into this:
>>
>> resource_management.core.exceptions.Fail: Execution of 'ambari-sudo.sh su hdfs -l
-s /bin/bash -c 'ulimit -c unlimited ;  /usr/hdp/current/hadoop-client/sbin/hadoop-daemon.sh
--config /usr/hdp/current/hadoop-client/conf start datanode'' returned 1. Error: JAVA_HOME
is not set and could not be found.
>>
>>
>> This seems to occur on all nodes when it attempts to start the services. Is this
a known issue?
>>
>> Thanks!
>>
>>
>>
>>
>> On Tue, Nov 24, 2015 at 3:05 PM, Naga Vijay <nagahive@gmail.com> wrote:
>>
>>> Hi,
>>>
>>> Was there an intermediate step for the host mappings?
>>>
>>> Thanks
>>> Naga
>>>
>>>
>>> On Tue, Nov 24, 2015 at 4:28 AM, cs user <acldstkusr@gmail.com> wrote:
>>>
>>>> Hi Sean,
>>>>
>>>> Many thanks for getting back to me. Uploading the cluster json as a
>>>> blueprint first, and then creating the cluster seems to have fixed it for
>>>> me :-)
>>>>
>>>> Thanks!!
>>>>
>>>> On Tue, Nov 24, 2015 at 11:28 AM, Sean Roberts <
>>>> sroberts@hortonworks.com> wrote:
>>>>
>>>>> You are correct that you need to:
>>>>> - 1. upload the blueprint to /api/v1/blueprints
>>>>> - 2. Then issue a cluster creation request to
>>>>> /api/v1/clusters/thenewcluster
>>>>>
>>>>>
>>>>> https://cwiki.apache.org/confluence/display/AMBARI/Blueprints#Blueprints-BlueprintUsageOverview
>>>>>
>>>>> Your existing cluster was likely not made from a blueprint so
>>>>> /api/v1/blueprints would be empty.
>>>>>
>>>>> --
>>>>> Hortonworks - We Do Hadoop
>>>>>
>>>>> Sean Roberts
>>>>> Partner Solutions Engineer - EMEA
>>>>> @seano
>>>>>
>>>>> On Nov 24, 2015, at 10:34, cs user <acldstkusr@gmail.com> wrote:
>>>>>
>>>>> Hi All,
>>>>>
>>>>> I've been looking at using blueprints to create a cluster. To do this,
>>>>> I have manually setup a cluster and then navigated to :
>>>>>
>>>>> :8080/api/v1/clusters/testcluster01?format=blueprint
>>>>>
>>>>> This has returned a json file which describes my cluster.
>>>>>
>>>>> I've then created a new cluster and I am trying to add the
>>>>> configuration via the json file from above, via a CURL post, with the
json
>>>>> as the data. The api is returning:
>>>>>
>>>>> {
>>>>>   "status" : 400,
>>>>>   "message" : "The properties [Blueprints/stack_version,
>>>>> Blueprints/stack_name] specified in the request or predicate are not
>>>>> supported for the resource type Cluster."
>>>>> }
>>>>>
>>>>> At the end of my json file, it contains the following:
>>>>>
>>>>>   "Blueprints" : {
>>>>>     "stack_name" : "HDP",
>>>>>     "stack_version" : "2.3"
>>>>>   }
>>>>>
>>>>>
>>>>> I'm submitting the post request to :
>>>>>
>>>>> :8080/api/v1/clusters/testcluster01
>>>>>
>>>>> Should this be working, or should I extract the minimum amount of data
>>>>> from the json file required to create my cluster and use this in a post
>>>>> request?
>>>>>
>>>>> I thought perhaps that before trying to create the cluster via the
>>>>> API, I would have to upload a blueprint (Perhaps the HDP blueprint
>>>>> mentioned above?). However, when I query the existing cluster, there
are no
>>>>> blueprints available
>>>>>
>>>>> 8080/api/v1/blueprints/
>>>>>
>>>>> {
>>>>>
>>>>>   "href" : "http://server.some.domain:8080/api/v1/blueprints/",
>>>>>   "items" : [ ]
>>>>> }
>>>>>
>>>>> Is it possible to capture the cluster config in this way from an
>>>>> existing cluster, setup via the UI, and setup a new cluster with it?
>>>>>
>>>>> Thanks!
>>>>>
>>>>>
>>>>>
>>>>>
>>>>
>>>
>>
>

Mime
View raw message