Puzzling filter in huge data

@filter in few data is a good use, but in huge data, there is a problem. Here is log:E1227 15:35:57.763290 11911 groups.go:848] Error in oracle delta stream. Error: rpc error: code = Canceled desc = context canceled

Would you provide more details of the issue that you are facing? It would be nice to see the query, how much data you have, dgraph version and example dataset.

Thanks very much for your reply!

There are 600 million nodes whose uid are start with 1 and end with 600 million, 300 million edges for “t”, 600 million edges for “a” and about 3 billion edges for “s”. It`s my schema:

t:int @index(int) .
a:[uid] @reverse @count .
s:[uid] @reverse @count .

My query is:


The problem happens when I use filter on “t” predicate.

BTW, The value of “t” predicate is equal to its uid.(1-600million)

Dgraph version : v1.1.0

Example data:

<2><t>"2" .
<2><a><2> .
<2><a><3> .
<2><s><144>(1=2,3=4) .
<2><s><240>(1=2,3=4) .
<2><s><226>(1=2,3=4) .
<2><s><519>(1=2,3=4) .
<2><s><9>(1=2,3=4) .
<2><s><265>(1=2,3=4) .
<2><s><525>(1=2,3=4) .
<2><s><367>(1=2,3=4) .
<2><s><35>(1=2,3=4) .
<2><s><477>(1=2,3=4) .
<2><s><72>(1=2,3=4) .
<4><t>"4" .
<4><a><4> .
<4><a><5> .
<4><s><276>(1=2,3=4) .
<4><s><559>(1=2,3=4) .
<4><s><473>(1=2,3=4) .
<4><s><280>(1=2,3=4) .
<4><s><108>(1=2,3=4) .
<4><s><560>(1=2,3=4) .
<4><s><186>(1=2,3=4) .
<4><s><390>(1=2,3=4) .

I solved that by changed my test-data to a set that the most num is not more than 3million. Maybe there is a bug when the num is too large.

1 Like

One more question, how long did the query take to return a response or fail with larger dataset?

It`s about 2 minutes

As far as I can see, this seems like a one off case. Does your query fail every time? 2 minutes is not a long enough time for bad things to happen. If it is acceptable for you to share the data and the queries with, I will be happy to look into it.

This topic was automatically closed 30 days after the last reply. New replies are no longer allowed.