ReadTs: X less than minTs: Y for key during query [Single Node setup] [v 1.0.15]

bug

(Rohan) #1

Apologies if this is a duplicate post, if it is then please forward me to the post where i can find the solution

I am getting this exception while my single node setup is up and running that to without any concurrent mutations happening.

I have loaded about 20M rdfs to the graph through bulk loading. i am using 1.0.15 version.

i have tried following :

  1. Retrying transactions. [DINT WORK]
  2. Removing zw directory and then restarting zero and alpha. [DINT WORK]
  3. bulk loaded and recreated data again. [ This way it had to work but not a preferable way]


(Michel Conrado) #2

You shouldn’t do it, Zero holds important information about the cluster.

Why not?


(Rohan) #3

Bcz it requires me to recreate my complete database. Which I don’t think should be the ideal solution.
However are you suggesting the creating graph db from scratch should be done when this error occurs ?


(Michel Conrado) #4

It’s perfectly normal for you to clear all paths and start from scratch.
However it is not recommended that you delete zw when you do the Bulkload.
That is a point.

I’m just asking, I didn’t suggest anything yet. Just said do not delete zw from original Bulk.

First of all you need to understand what is triggering this error. What causes the error is more important now than anything. I think maybe it’s related to zw. But otherwise it is necessary to understand every context step by step.


(Rohan) #5

Hi Michel,
First of all thanks a lot replying. Much appreciated.
I just wanted to know where and why this problem occurs since I am really new to Graph Databases.

Which all path should I clear ? and what’s the ideal to restart from scratch ?


(Michel Conrado) #6

1 - Clean these paths.

/${yourPath}/dgraph/p
/${yourPath}/dgraph/w
/${yourPath}/dgraph/zw

2 - Start Zero.
2.1 - Start Bulkload.

3 - Keep Zero (zw) from the Bulkload.
3.1 - Start your cluster.