How can we import NULL value for numeric type from CSV file using Nebula Importer in version 3.2 ? https://github.com/vesoft-inc/nebula-importer/issues/168

Hi all. I'm trying to display all the nodes and edges using the following query MATCH (m)-[e]->(n) RETURN m,n,e limit 10; But I'm getting the following error: Scan vertices or edges need to specify a limit number, or limit number can not push down Can someone please clarify?

hello! what's the recommended migration setup for ingesting larger datasets (say SNB benchmark with bigger SFs - 1k-30k)? specifically how the migration should be handled to make it efficient, should I use exchange or importer? if exchange, is hdfs a must and would migrating to ssts first be fastest?

Hello! What's the best/fastest way to fetch large number of node indices, say 50M-100M? These would be fed as starting points to other queries. Using a simple

Match

returns 5M indices in about 5 minutes.

What are indices, are those Vertices ?

I'm testing on SNB SF100 generated with

nebula-bench

. Graph is deployed via docker-compose running on a 64 vCPU machine with 256GB RAM. I add a tag index to Post and rebuild:

CREATE TAG INDEX IF NOT EXISTS post_index on Post()
REBUILD TAG INDEX post_index;

After that i use

lookup

. Querying for Posts (~60M instances), this query returns in around 5 minutes:

LOOKUP ON Post yield id(vertex) | yield count(*)

whereas

LOOKUP ON Post yield id(vertex)

stalls. I tested both using console as well as python client. Am I missing something? How should I retrieve large number of results then? Use vertex scan? Some kind of pagination?

Hi team. I'm trying to model a graph in Nebula and currently stuck in modelling a hypergraph. The problem I'm dealing with is something similar to this question (https://stackoverflow.com/questions/61395285/graphdb-node-as-a-property-of-a-relationship) There are two answers here. 1. First answer suggests to break the hypergraph by introducing an intermediatory node 2. Second one suggests to add an edge property With respect to the Nebula Graph, which approach would be better here (Considering the performance for large datasets and scalability)

hi, I'm having trouble understanding some of the Grafana metrics we have a single Nebula importer instance running a history load to our cluster (v3.2.0, 8 nodes). the importer logs a steady rate of ~250k rows/s:

2022/09/13 08:05:02 [INFO] statsmgr.go:89: Tick: Time(290.00s), Finished(71750768), Failed(0), Read Failed(0), Latency AVG(1887us), Batches Req AVG(2043us), Rows AVG(247416.41/s)
2022/09/13 08:05:07 [INFO] statsmgr.go:89: Tick: Time(295.00s), Finished(72978112), Failed(0), Read Failed(0), Latency AVG(1887us), Batches Req AVG(2043us), Rows AVG(247383.40/s)
...

at the same time, a Grafana metric

nebula_storaged_num_add_vertices_rate_60{instanceName=~"$hosts"}

is reporting ~110k rows/s. could someone help interpret this metric pls? thank you!

Hi Team. Say suppose I'm using NebulaGraph v3.2.0 as tar binary file. How should I upgrade to v3.2.1. When I checked the official documentation, there are only upgradation procedures for the database installed through source compilation or RPM/DEB package. Can you please elaborate on how to upgrade when using tar package?

Hi team. Is it possible to mention domain instead of IP for meta_server_addrs in configuration files?

I think no. But you can try.

Does the Nebula Java client supports transactions (commit, rollback) ? I checked the client but unable to find any transaction support. Kindly help me in this regard

No transactions, just atomic read/writes using raft.

May i know the reason behind not supporting transactions in Java client? Aren't transactions necessary? In fact, one of use-cases demands it.

For now, we have to do read after write and do rollback if needed from the application layer as NebulaGraph DB core doesn’t support transactions for now.

Okay. Thanks for the information.

One more query. I'm using GO statement to retrieve 2nd degree vertices and the result set contains vertices mapped with either of the two tags (Tag_A or Tag_B). In such a case, how do I apply filter/condition to print only the vertices of Tag_A type? GO 2 STEPS FROM "123456" OVER * WHERE tags($$)=="Tag_A" YIELD id($$); I tried the above query but couldn't achieve the desired result

Hi. Does Nebula Java client supports batch insert?

Can anyone please confirm if the Nebula Java client supports batch insert? I checked and looks like it is not supported. Wanted to confirm.

Hi team. Do I have an advantage in terms of performance/lookup speed when choosing int64 as VID over string? I've a vertex which stores a domain name in it (The vertex is associated with a tag named "domain" and domain name is one of its properties). In terms of querying, domain name will be used in most of the cases to query a particular vertex. In such a scenario, should I choose domain name as VID (String type) or use hash algorithm such as snowflake (As suggested in the docs) to generate an int64 and use it as VID. Can you please clarify on the query performance part when using int64 over string as VID?

Any plans to provide support for rolling upgrade? Upgrading clusters with zero-downtime is necessary for us.

It can be done for minor version upgrade, eg 3.2.0 to 3.2.1 (will se for 3.3.0 but I suppose it will be the same) which I did few days ago, upgrading cluster node by node without downtime.

Oh okay. Can you please confirm if the support for rolling upgrade (For major version) is in your roadmap? @wey

Hi team. I've got few queries regarding the usage of domains(Instead of hosts IP) in nebula 1. I had setup 3 meta nodes and DNS mapping was made for 3 of the hosts IP to a single domain. When I configured the domain in place of meta-addrs in the conf file, only one of the IPs were mapped as meta and not all three. May I know the reason for this behaviour? 2. In client side, from Java, what is the suggested method to configure graph processes' IPs? As hosts IP may change at times, i thought of having a DNS mapping to the IPs and provide the domains in Java while making connection. Is this method a suggested one? Also, are DNS lookup results cached in the client side and if cached, how frequent they're invalidated and updated in client side? It would be great if you guys share some insights on this matter. Thank you

I’m using IP addresses in config files, because it is at the end, stored in internal structures of Nebula database. Connecting with DNS names can be used, so client connections can be with DNS names specified.

What this statement means, from documentation for INSERT EDGE ?

When inserting an edge that already exists, INSERT VERTEX *overrides* the edge.

I’ve executed some queries using FETCH with VertexID but for first request for specific VertexID always got

[ERROR (-1005)]: Storage Error: Not the leader of 103. Please retry later.

second request returns valid response. Also some request don’t return any results for keys that were inserted earlier. I’ve upgraded my cluster to 3.3.0 version from 3.2.1 one node at a time. fyi: one parameter for rocksdb I’ve changed

--rocksdb_compression_per_level=no:no:lz4:lz4:lz4hc:lz4hc:lz4hc

from default setting that was empty string can that change have impact to not be able to query some of the data ? I’ve actually changed another one, using query optimizer, previously it was disabled !

--enable_optimizer=true

Did you tested compression change on already populated database and is there need to execute some compaction or some other action to have data available for query ?

Here are some errors I’ve got when executing LOOKUP query

E20221129 23:24:55.530879 332829 StorageAccessExecutor.h:39] IndexScanExecutor failed, error E_RPC_FAILURE, part 90
E20221129 23:24:55.530885 332829 StorageAccessExecutor.h:39] IndexScanExecutor failed, error E_RPC_FAILURE, part 92
E20221129 23:24:55.530894 332829 StorageAccessExecutor.h:39] IndexScanExecutor failed, error E_RPC_FAILURE, part 95
E20221129 23:24:55.530901 332829 StorageAccessExecutor.h:39] IndexScanExecutor failed, error E_RPC_FAILURE, part 96
E20221129 23:24:55.530913 332829 StorageAccessExecutor.h:136] Storage Error: part: 124, error: E_RPC_FAILURE(-3).
E20221129 23:24:55.530952 332825 QueryInstance.cpp:137] Storage Error: part: 124, error: E_RPC_FAILURE(-3).