Slack discussion Digest

[wey-gu]

Weekly Digest of Nebula Graph Slack Chanel discussions will be shared in this thread.

This is not needed, please visit https://community-chat.nebula-graph.io or https://www.linen.dev/s/nebulagraph/ to see all threads.

[wey-gu]

How to generate SF100 data properly w/o suffering from running out of memory

• Use https://github.com/ldbc/ldbc_snb_datagen_hadoop to generate data.
• Increase the memory limit. i.e. export HADOOP_CLIENT_OPTS="-Xmx100G"

Solution provided by @HarrisChu

[wey-gu]

E_WRITE_WRITE_CONFLICT during Data balance

It's reported by Marija in Slack that E_WRITE_WRITE_CONFLICT: -3073 will be encountered during data balance remove in 3.2.0.

It is related to TOSS(transaction over storage side), which means a transaction for an edge during data mutation, with the experimental flag being set to true this feature was enabled, too.

@wey-gu suspects that when data balancing (remove) is in progress, there will be edges in the involved partition(as leader) being locked, before the leader is changed, thus it’s causing this error when corresponding edge data is being manipulated/written.

#4453 was reported to enable TOSS to be disabled for experimental flag enabled(for data balancing feature switch).

[wey-gu]

NULL value handling in Nebula-Importer

This was discussed and with mitigation provided by @goranc before vesoft-inc/nebula-importer#168 was solved, we could preprocess CSV file to use NULL standing for NULL, which is a golang behavior/feature.