JanusGraph index stuck in "INSTALLED" state after running REGISTER_INDEX command

[Ravindra Gupta]

I'm trying to create a composite index on the title property in JanusGraph using the Gremlin console.

Here's what I've done so far:

 1. Created the `title` property key.
 2. Built a unique composite index on this key named `byTitleUnique`.
 3. Waited for the index to show in the INSTALLED state using

>`mgmt.getGraphIndex('byTitleUnique')`

 4. Attempted to update the index status to REGISTERED using the following commands:
```

mgmt = graph.openManagement()
mgmt.updateIndex(mgmt.getGraphIndex('byTitleUnique'), SchemaAction.REGISTER_INDEX).get()
mgmt.commit()

```
After running the command and waiting a long time (even over an hour), the index status still remains in the `INSTALLED` state.

As per my understanding, after this step, I should be able to **Reindex** and then **Enable** the index. But since the index isn’t moving past the `INSTALLED` state, I'm stuck.


----------

Questions:
----------

 - Why isn't the index transitioning from `INSTALLED` to `REGISTERED` after
   running the command?
   
 - Is there anything else I need to do to make this transition happen?
   
 - Could this be related to the backend (I'm using Cassandra and
   Elasticsearch)?

Any help or insights would be appreciated.

[Andrea Child]

Hello,

 

I do not use JanusGraph but I noticed that according to JanusGraph index documentation described here https://docs.janusgraph.org/schema/index-management/index-performance/ the user is expected to wait until the index becomes registered and is not expected to set the state to registered.

 

//Wait for the index to become available (i.e. wait for status REGISTERED)

ManagementSystem.awaitGraphIndexStatus(graph, "myIndex").call();

 

I was going to provide a link to the JanusGraph github discussions and noticed that you have already opened a corresponding one https://github.com/JanusGraph/janusgraph/discussions/4792. I hope the JanusGraph folks can help you there.

 

Cheers,

 

Andrea

 

--
You received this message because you are subscribed to the Google Groups "Gremlin-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to gremlin-user...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/gremlin-users/16a0ff03-d7a9-4d3c-965d-d52786d56bb6n%40googlegroups.com.

[Bernard Dudley]

Hi,

I'm still relatively new at this myself, but I have suffered from this issue and have a vague memory of what went wrong, and our solution, in our case.

My understanding is that (in our case, yours might not be the same) JanusGraph is waiting for the index request to propagate to all attached instances. We have a scheduled backup procedure that spins up an extra instance of JanusGraph in order to create the backup, and if this instance is not shut down gracefully, it is somehow still 'attached' and therefore expected to respond to the propagation of the index. Since it never responds to the primary (because it's gone away), the status of the index does not progress.
I cannot remember whether our solution was some clever way of detaching the ghost instance, or we simply restarted the primary instance (I suspect it was the latter).  

I hope this helps,

Bernie

To view this discussion visit https://groups.google.com/d/msgid/gremlin-users/MW3PR13MB391330CE651DFB897C5651A69365A%40MW3PR13MB3913.namprd13.prod.outlook.com.