error during query: readTs value less than minTs value

[mpires]

• What version of Dgraph are you using?
  1.0.4 nightly (98c6102), also tried releases 1.0.3 and 1.0.4

• Have you tried reproducing the issue with latest release?
  I have, with the same erroneous results.

• What is the hardware spec (RAM, OS)?
  3 GCE instances (4 vCPUs, 20 GB, 30GB SSD)

• Steps to reproduce the issue (command/config used to run Dgraph).
  A fresh dgraph installation with the p folder that I sent privately to @pawanrawal
  And the following query:

{ranges(func:eq(view,"c4c97e30-87b1-11e7-bf97-a71a48b50a2d")) { xid }}

• Expected behaviour and actual result.
  While some queries do work, the predicate view for certain keys (such as the above) seems corrupt, giving out the error readTs: <X> less than minTs: <X+Y>

[pawanrawal]

Do you have the logs from the three Zero servers by any chance?

[aditya-pr]

getting same error after increased mutation rate for a while.

[manishrjain]

This is a temporary error. A transaction retry should fix this.

[akshaydeo]

This is not temporary!And it does not go away with a retry as well.

This goes away, once the mutations stop happening on the entities which you are trying to read. This again technically means, transactions built in dgraph are not working the way they should.

Could you please look into it!

[willcj33]

I am getting this issue as well. Just wanted to chime in. Its when I have a very high load of mutations flowing though.

[manishrjain]

Did you try this with v1.0.5? There was a bug around that which got fixed in v1.0.5.

Just to clarify: Under a high load, this could happen. Because many writes are going into one list, and the list gets folded rapidly as it grows. Every time the fold happens, any txn trying to read from below that fold gets this error. But, it should go away as mutation load decreases.

The bug that was fixed caused this to happen even when no writes were going on. If you still see that, holler!

[willcj33]

Yes, I am on 1.0.5. It's definitely under high load. Is it something I can just retry after a backoff? I mean ive written like 100k events which is a alot of mutations per single event, in less than 1 hour, and I only got this error like twice. Also, we don't have alot of ram or cpu on our 5 machines we are testing with right now, if that matters.

[manishrjain]

Yeah, it's a temporary error. You can just retry with a backoff.

All: This should be fixed with v1.0.5. If you're still seeing this error with little mutation load, ping on this thread. Otherwise, I'll close it in a few days. (This error happening temporarily is perfectly normal with mutations going on in the system)

[manishrjain]

@akshaydeo : Can you retry from master, and check if you still see this error? Again, seeing this temporarily under high load is OK, what we're looking for is if you can't get out of this error even after the mutations stop.

[akshaydeo]

This works for me now. I had to add a few retries in case of failures.

[manishrjain]

Cool. Closing it then.