Details
-
Bug
-
Status: Closed (View Workflow)
-
Major
-
Resolution: Fixed
-
1.2.1
-
None
-
None
-
2020-1, 2020-2, 2020-3, 2020-4, 2020-5, 2020-6, 2020-7
Description
During postConfigure for a new cluster, we are choosing external storage for our Performance Modules in order to enable HA failover reattachment of EBS volumes to alternate PMs in the event of a PM failure.
Whether one chooses to allow postConfigure to create new EBS volumes for the PM dbroots or create the volumes their self and provide the volume IDs, postConfigure incorrectly assumes the devfs node for the dbroot to be a legacy Xen device descriptor (such as /dev/xvdg, /dev/xvdh, etc.) and writes these incorrect values to /etc/fstab.
Attachments
Activity
This is a very thorough description of both the problem and the solution! Much appreciated!
I talked with Todd, and since we're trying to be more platform agnostic, the decision was to strip out the AWS-specific features in the system. Users will configure their machines the same way whether it is local, or in cloud XYZ. That will simplify documentation, support, and the code. As a side-effect, that removes the feature of CS manipulating the fstab file.
In the meantime, customers can configure their clusters by answering 'n' to the question "Do you want to have ColumnStore use the Amazon AWS CLI Tools [y,n] 
> " in postConfigure, and following the non-AWS instructions otherwise.
Build verified:
1.2.6-1 BB-20200502
1.4.4-1 Jenkins-20200506
1.5.0-1 BB-20200507
All three releases above no longe have the option to use AWS CLI. Customer will have to install ColumnStore on AWS instances just like installing on standalone Linux OSs.
Another customer reported this issue