Chunk size error


#1

I have an array with one attribute and two dimension like shown below:

connect<e:uint64 null default null>[src=0:,65536,0, des=0:,65536,0]

This array contains data about 5gb. When I join this array with itself it gives following error:
Error id: scidb::SCIDB_SE_METADATA::SCIDB_LE_LOGICAL_CHUNK_SIZE_TOO_LARGE
Error description: Metadata error. Chunk size should not exceed 2^63.

My query is:
select * into test from connect as c1 join connect as c2 on c1.des=c2.des and c1.src!=c2.src and c1.src>c2.src;

Anything that can be done?


#2

Hey:

You should use AFL as AQL is not as robust yet.

  1. Try
    iquery -aq filter( cross_join(connect as c1, connect as c2, c1.des, c2.des), c1.src != c2.src and c1.src > c2.src)

  2. You can also try the equi_join operator: https://github.com/paradigm4/equi_join


#3

Hello, thank you for the answer.

Chunk size error didn’t show up but I did get new error.
The error when running the above query is:

SystemException in file: src/network/BaseConnection.h function: receive line: 414
Error id: scidb::SCIDB_SE_NETWORK::SCIDB_LE_CANT_SEND_RECEIVE
Error description: Network error. Cannot send or receive network messages.

The “connect” array has almost 4 million data and SciDB is configured in a single machine with 4Gb Ram.
Any help for this kind of error. I cannot figure out why network issues occured in a single machine.