ugurtosun
Node Link

Hello everyone,

I am using community edition 4.3.3. I started a long-time running query (graphsage training) from browser. Then my browser closed. When I am checking with "top" and "htop", process is still running. It is good, but there is way list running queries in the server ?

In addition is it possible to see query history ?

Thanks.

Comments
Cobra
Ninja
Ninja

Hello @ugurtosun and welcome to the Neo4j community

I guess you are looking for CALL dbms.listQueries().

For the history, you can use :history

Regards,
Cobra

ugurtosun
Node Link

@Cobra thanks a lot. In addition, I got the result that second row (graphsage train) is process investigated by me. I started yesterday, but still not completed. Is it normal for a graph consisting of (82,339,589) nodes and (219,521,164) edges.


Cobra
Ninja
Ninja

It is a big graph

Have you allocated a good amount of RAM for the database?

Can we see the query and the configuration of GraphSAGE?

ugurtosun
Node Link

Thanks a lot @Cobra . I am sharing the full config file becuase I am newbie for neo4j. I am open to any kind of suggestion to configure system sources. I have 48 core cpu (but community edition works for 4) and 500gb total RAM in the system (Normally I set 250gb heap memory in the config file, but now I comment out this config because of using all resources. I guess it is not wrong ).

query:

CALL gds.graph.create(
'derin',
{
Transaction: {
label: 'Transaction',
properties: ['trx_amount']
}
, Customer: {
label: 'Customer'
}
, GL_Branch: {
label: 'GL_Branch'
}
, Scenario: {
label: 'Scenario'
}
, Withdraw: {
label: 'Withdraw',
properties: ['trx_amount']
}
}, {
SENDS: {type: 'SENDS', orientation: 'UNDIRECTED'},
RECEIVES: {type: 'RECEIVES', orientation: 'UNDIRECTED'},
INSPECTS: {type: 'INSPECTS', orientation: 'UNDIRECTED'},
SERVING: {type: 'SERVING', orientation: 'UNDIRECTED'},
WITHDRAWS: {type: 'WITHDRAWS', orientation: 'UNDIRECTED'}
}
)


CALL gds.beta.graphSage.train(
'derin',
{
modelName: 'derin_model',
featureProperties: ['trx_amount'],
embeddingDimension: 16,
projectedFeatureDimension: 1
}
)

neo4j-config.txt (17.2 KB)

Cobra
Ninja
Ninja

You should have a look at neo4j-admin memrec like in this post to know how to allocate ram.

For the train, you should test your configuration on a sample to make sure it is what you want. Since this is a large dataset, I think it is normal for training to take time.

Check also default values of the configuration, maybe you can play with some to reduce the time it takes.

ugurtosun
Node Link

memrec command very informative. Thanks.

Version history
Last update:
‎12-02-2021 12:38 AM
Updated by:
Contributors