Indefinitely growing homepage.sr_index_docs table and forgotten search nodes

Created:
Last Update:

Author: Christoph Stoettner
Read in about 2 min · 262 words

Work From Home

Photo by Benjamin Hershey | Unsplash

The last weeks I had twice the issue of a huge homepage database with HCL Connections. Both systems weren’t installed by me, but I reviewed them.

So I know this issues quite well since I faced it the first time years ago. This always happens when you migrate Connections environments side-by-side, and you forget to remove the old search nodes.

sr_index_docs is used to make all search nodes indexing newly upload files. When each search node has indexed the file, the entry gets removed from the database. As the old nodes are no longer running, they can’t index, and so these entries never get deleted.

In larger environments the table grows to millions of documents and the database size grows from some GB into the hundreds of GB. You see this growth mainly during database backup, or when you want to move the data to a new machine. So often nobody recognizes the huge waste of space.

The documents in sr_index_docs disappear as soon as the old search node entries get deleted.

Run wsadmin.sh from Dmgr01/bin directory:

execfile("searchAdmin.py")
SearchService.listIndexingNodes()

This will show you a list of search nodes:

screenshot listIndexingNodes

You can identify the old nodes on the timestamp of the Last crawl version, or with the Node & Server name. Now run the following command with the old search node names:

SearchService.removeIndexingNode("Node01:cluster1_server1")

There are two knowledge base entries describing these phenomena, but they are a bit short and links / solutions are missing.

Author
Suggested Reading
Card image cap

I created a git repository with some smaller CSS files to fix some annoyances within HCL Connections.

I started with this to prevent Orient Me to load fonts from external URLs or Elasticsearch Metrics to break the UI on larger screens. These issues are solved after the last updates I got from support, but Blogs and Tailored Experience Wizard can be improved with some simple rules.

Created: Read in about 3 min
Card image cap

I showed, in several slides and sessions, how you can use the search-admin role in the search application of HCL Connections for troubleshooting and reviewing some key configurations.

In several environments, my user or other administrative users have this role, just to access the link to /search/serverStatus for example.

Created: Read in about 3 min
Card image cap

Elasticsearch in HCL Connections Componentpack is secured with Searchguard and needs certificates to work properly. These certificates are generated by bootstrap during the initial container deployment with helm.

These certificates are valid for 10 years (chain_ca.pem) or 2 years (elasticsearch*.pem) and stored in the Kubernetes secrets elasticsearch-secret, elasticsearch-7-secret. So when your HCL Connections deployment is running for 2 years, the certficates stop working.

Created: Read in about 3 min