How to config number of retry attempts and the retry delay in Janusgraph

[dinhphu2k1-gif]

I'm newie in JanusGrahGraph.I use Hbase as storeage backend ans Elasticsearch as index backend.

When i read transaction fail, it mentions that

The number of retry attempts and the retry delay are configurable
But i don't find it in configuration. Can anyone help me? Thank everyone!!

[vtslab]

Good question, the documentation seems to be ambiguous here!

I believe the remark about configurable retries refers to the storage.lock.retries and storage.lock.wait-time properties. These are relevant for transactions on elements with a uniqueness constraint, see:
https://docs.janusgraph.org/v0.4/advanced-topics/eventual-consistency/#data-consistency

If you use JanusGraph Server, also relevant is the evaluationTimeout mentioned in https://tinkerpop.apache.org/docs/current/reference/#_configuring_2 and the responsibility of the client for retries in https://tinkerpop.apache.org/docs/current/reference/#request-retry

[FlorianHockmann]

As @vtslab already pointed out, transactions shouldn't matter here.

One question from my side: The log message you pasted shows that your addV() also includes a lot of property() steps, do you also get the timeout without these property() steps, or only with them? And if it only occurs with the property() steps: Do you have a unique index defined on one of these properties?

[dinhphu2k1-gif]

@FlorianHockmann i really don't know it occurs with the property() step or only with them. After using g.addV(), i use many property() steps, finally using next() step.

And i have build a mixed index with these properties

[FlorianHockmann]

Not sure but this could be caused maybe by problems with your mixed index backend. I would try the traversal with a much reduced number of property() steps to see whether the problem still occurs and ideally with property keys for which you don't have a mixed index.

[dinhphu2k1-gif]

@FlorianHockmann did you find any problem with mixed index?

[FlorianHockmann]

did you find any problem with mixed index?

No, not really. I just want to get a better idea of what could cause the transactions to be too slow here apparently and I think mixed index modifications could be something that's slowing down the transaction.