topic Re: How to Re-Connect the Installer with the Cluster in HPE Ezmeral Software platform

How to Re-Connect the Installer with the Cluster

msaidbilgehan — Tue, 08 Aug 2023 07:52:14 GMT

After failed incremental installation at the cluster, I abort the progress and the installer reset itself. After reset, Installer goes back to installation page. Now I need to re-connect the installer to the cluster.
Checked the files "/opt/mapr/installer/data/groups.json" and "/opt/mapr/installer/data/last_known_state.yaml" bot have old cluster information.

I tried to import the last know state (last_known_state.yaml) at Installer UI but not worked. What should I do for recover old cluster information?

Re: How to Re-Connect the Installer with the Cluster

Shishir_Prakash — Tue, 08 Aug 2023 19:26:19 GMT

Can you please explain a bit more about this issue?

Re: How to Re-Connect the Installer with the Cluster

msaidbilgehan — Thu, 10 Aug 2023 06:24:57 GMT

@Shishir_Prakash

Well, let me start from scratch.

I installed mapr installer using the mapr-setup script of Data Fabric to my one node named "mapr-installer"
I installed data fabric using mapr-installer UI to 3 different nodes named "node-[1-3].example_domainname.com" with one ZK and CLDB. ZK and CLDB placed a node named "node-1.example_domainname.com"
After the installation of the Data Fabric over 3 nodes, I used incremental installation to add more ZK to the cluster. So I select the other 2 nodes to install ZK too.
Installation Failed (Here is the ticket about details of installation. But this issue is about Installer, the installation fail ticket is about ZK. So different cases happened at the same time)
After fail happened at Installer, I selected Abort the installation to recover the old state (which is a clean installation with one ZK node)
Abort action resets the Installer and now Installer not seeing the cluster already running. Now Installer is at the setup page and like there is no node installed Data Fabric. But Nodes are running services already.
I tried to recover the Installer from "import state" option but not worked. As I mention above, I checked both "/opt/mapr/installer/data/groups.json" and "/opt/mapr/installer/data/last_known_state.yaml" files and both contains old successfully installed state of cluster.

I guess every step is clearly described. Please don't hesitate to ask questions because it may happen anytime and I want to know how to fix it?

Re: How to Re-Connect the Installer with the Cluster

GFustagni — Fri, 25 Aug 2023 10:35:06 GMT

Installer DB can be regenerated using the 'Probe' and 'Import' Stanza commands as documented here:

https://docs.ezmeral.hpe.com/datafabric-customer-managed/73/AdvancedInstallation/Stanzas/UsingProbeImport.html

Re: How to Re-Connect the Installer with the Cluster

msaidbilgehan — Fri, 01 Sep 2023 07:14:34 GMT

@GFustagni

I guess here is the problem;

./mapr-installer-cli import -n -t /tmp/stanza.yaml
HPE Data Fabric Installer SDK
Logging in to localhost
Are you sure you want to import cluster (your current installer state will be lost)? [y/n]:y
Are you sure you want to reset config? [y/n]:y
Resetting the cluster...
Done
WARNING: You have configured your cluster with 2 data node(s). MapR does not support clusters with less than 5 data nodes in production outside MapR Edge. A data node is a node that runs the MapR FileServer process and is responsible for storing overall cluster data. Clusters with less than 5 data nodes will have a lower failure tolerance, and may not be able to sustain data integrity in the case of more than one failure. If this is a non-edge production cluster, please add more data nodes.
WARNING: You have configured your cluster with 1 cldb node(s). MapR does not recommend clusters with less than 3 cldb nodes in production. A cldb failure in a cluster with less than 3 cldb nodes can result in a cluster that is unaccessible or in read only mode. If this is a production cluster, please add more cldb nodes.
ERROR: import command failed
These hosts [node-2.treo.com.tr] must have 'mapr-core' in one of the 'groups' section of the template
Log files mapr-installer.log and installer_cli.log can be found at /opt/mapr/installer/logs

But in the community edition, CLDB and Zookeeper services are forbidden to install more than one node. How can we solve this issue? We should import the last known state to the installer even if the cluster has one node Zookeeper and CLDB installed.

Here are the installer logs: https://drive.google.com/file/d/1qYwJn3591bKNIPlRlW_Un9KFy7e7F3fx/view?usp=sharing