ambari-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Clark Breyman <cl...@breyman.com>
Subject Re: Creating a second cluster via REST api corrupts ambari configuration
Date Sat, 27 Jun 2015 01:50:39 GMT
Thanks Sumit. https://issues.apache.org/jira/browse/AMBARI-12181

The following steps seemed to have gotten the cluster back operational.

delete from clusterstate where cluster_id <> 2;
delete from alert_grouping where definition_id in (select definition_id
from alert_definition where cluster_id <> 2);
delete from alert_definition where cluster_id <> 2;



On Fri, Jun 26, 2015 at 5:43 PM, Sumit Mohanty <smohanty@hortonworks.com>
wrote:

>  While it is possible the data model is indeed corrupt but more than
> likely there are sections of code that stop at the first cluster they see
> and that is creating some confusion. I think you can delete the
> second/third cluster from the cluster* tables and after a restart Ambari
> Server should get back in shape.
>
>
>  Tables to check ...
>
>
>  * clusters
>
> * clusterstate
>
> * cluster_version
>
> * clusterservices
>
> * clusterconfig
>
> * clusterhostmapping​
>
>
>  Also, please go ahead a create bugs based on your observation.
>  ------------------------------
> *From:* clarkbreyman@gmail.com <clarkbreyman@gmail.com> on behalf of
> Clark Breyman <clark@breyman.com>
> *Sent:* Friday, June 26, 2015 5:18 PM
> *To:* user@ambari.apache.org
> *Subject:* Creating a second cluster via REST api corrupts ambari
> configuration
>
>  While experimenting with the REST api, I attempted to create a second
> cluster via POST to /api/v1/clusters/test. The result was having the
> service is a corrupted state as follows:
>
>  - Clusters lists the second cluster I created with the status "cluster
> creation is in progress". The original cluster and the third attempts were
> not listed.
> - All three clusters are listed in the popup associated with the Versions
> "Install on" button.
>
>  I understand (now) that Ambari does not yet support multiple clusters
> but it seems odd that the data model can be corrupted by an API operation.
>
>  Is there any way to fix my database without scraping and reprovisioning
> the entire cluster?
>

Mime
View raw message