Install GDS Library on Neo4j AuraDB GCP Professional Tier

[yazan_akkad95]

Hello Everyone,

I am using Neo4j Aura Professional Database as-a-Service on GCP. I need to use the GDS library but I can find it preinstalled. I went to the documentation for installation guide, but I found multiple options for installation and I am not sure which one I should use.

I hope you can guide me.

[alicia_frame1 · Accepted solution]

The Graph Data Science library is not available on AuraDB Professional, and it's not possible for users to install custom plugins.

If you're looking for GDS, you can sign up for the early access program to AuraDS: AuraDS for Data Science - Neo4j Graph Data Platform

[Reiner]

Hi Alicia,

thanks for your answer. I would have two more questions on this:

1.) Can GDS be used on any Neo4j-clusters outside of Aura? (local, Azure, ...)

2.) When using AuraDS and the GDS-use-case should work with real-time data from productive db in Aura and should write back results into relations in productive db - how can this be handled? (e.g. a kind of a skill-matching algo) (same for any cluster if your answer on 1. is "no") (please don't tell "batch import/export of data" - exactly that would not be what I would expect from Neo4j).
(I'm asking about Enterprise Tier)

[yazan_akkad95]

Hi Alicia,

Thank you for your valuables input. Then I'll be waiting for AuraDS impatiently since I'm working on entity resolution using Neo4j and I need the GDS library.

Regarding the single instance solution, I've already did that and I noticed that the GDS library is restricted. I checked $HOME_NEO4J/plugins and upgraded the library from 1.6 to 1.8. I also applied the required modification to the neo4j. conf file. However, I'm stuck at the next step which is restarting Neo4j. Do you mean Stopping the VM instance and Starting it again? When I do that, the changes I applied to the conf file are reset back to its original form.

[alicia_frame1]

Hi @yazan.akkad95 how are you deploying Neo4j? Is it from one of our marketplace images, docker, or you've pulled something from our download center and set it up yourself?

[yazan_akkad95]

Hi Alicia,

I got it as an image from the marketplace. But I think I found them mistake. I was applying the changes to the neo4j.conf file in var/lib/neo4j/conf rather than etc/neo4j/neo4j.conf

I think that's where I was doing it wrong.

[alicia_frame1]

1. GDS is, generally, intended for single instance deployments of Neo4j. Because of how GDS runs, and it's resource utilization, it can cause cluster instability when run on a core member. However, you can run GDS on a read replica. Clusters aside, you can run GDS on desktop, local installations, docker, cloud images and more. Full details here: Installation - Neo4j Graph Data Science

2. There are a few different options, depending on your data volumes, use case, etc. The easiest answer is that we're developing clone functionality between AuraDB and AuraDS, planned for release in the next few months. More complicated solutions include using Kafka to sync data (Enabling your business with the power of connected data | by Maruthi Prithivirajan | Jan, 2022 | Med...) , or using our drivers or other ETL tooling. You could also deploy yourself where you can have less typical configurations - like read replicas, etc.