Pipeline fail: ER_DISTRIBUTED_NOT_MASTER message


#1

I have installed MemSQL Studio v1.5.2 on the master of a 4-node v.6.7.10 cluster (master, child aggregator, and 2 leaves). After clicking Pipelines I have the following error displayed:

ER_DISTRIBUTED_NOT_MASTER: This instance is not the master aggregator. information_schema.PIPELINES is not permitted on child aggregators and leaves.

Any pointers on how to resolve the issue?


#2

Hi daemon (cool name by the way). Thanks for asking this question!

Two questions to help me understand your issue:

  1. What method did you install MemSQL in – AWS, Docker, or Linux?
  2. Are you sure you’re on the master aggregator and not the child aggregator?
    2.1 If you are on the master aggregator and you are getting that message, that is a little weird. But if you are NOT on the master aggregator and trying to use Pipelines, then please try again on the MA.

#3

Hi Jacky,

  1. The cluster is in AWS but it was a manual setup - i.e. I didn’t use the Cloudformation script but followed the on-premise guide. Eventually this will be part of Terraform and Ansible for our automated deployments.

  2. I am indeed on the Master.

Please let me know if you need any additional information.


#4

Hey Jacky, Mahir was able to help me get this resolved by creating a new cluster within studio.


#5

Hi Jim,

Glad we were able to get this resolved. For anyone else following this thread - it seems that studio was actually connected to Child Agg instead of MA. We confirmed that and after connecting to MA, pipelines module started working. Additionally if you’d like to delete the faulty cluster entry from Studio UI, here are the instructions:

  1. Open this file in the editor: vi /var/lib/memsql-studio/studio.hcl

  2. There should be two clusters in this file. Go ahead and delete the first one that’s connected to the CA. File with one cluster should look like this:

    version = 1

    cluster “7f198a94-dcaf-420e-a432-6da1d70a1e46” {
    name = “MiB Cluster”
    description = “Test”
    hostname = “”
    port = 3306
    profile = “PRODUCTION”
    }

  3. Restart memsql-studio service and the other cluster should be gone.


#6

Thanks @mbhojani for helping, and I’m glad you could get this figured out @daemon_of_chaos!