Can I have conflict in Riak with using a vclock getting by a get

Question

I want to know if I can have conflict in this scenario :

#!/usr/bin/env bash

curl -XPUT -d '{"bar":"baz"}' \
     -H "Content-Type: application/json"  \
     http://127.0.0.1:8098/riak/obj/1

response=$(curl -I http://127.0.0.1:8098/riak/obj/1 | grep 'X-Riak-Vclock:' | egrep -o ' .*$')

curl -v -XPUT -d '{"bar":"foo"}' \
     -H "Content-Type: application/json"  \
     -H "X-Riak-Vclock: $response" \
     http://127.0.0.1:8098/riak/obj/1

In some words :

First I have no object for the key 1 I put the {"bar":"baz"} value with the PUT of the http api.

Then, I read the value with a get. And I store the vclock in variable.

And finaly I put a new value {"bar":"foo"} for the key 1

Is there a case where I can have {"bar":"baz"} for the key 1 ? If Riak has a conflict, it will be resolve with vclock ?

Thanks !

Answer 1

It depends how your Riak database is configured, either globally or if you changed the default configuration of the bucket you're using. If you keep the default config, your second PUT (with the vclock) might: - fail, if someone updated the key behind your back (rare), and the vclock data you have is already obsolete. You need to re-read the value and update it. Best is to have a retry mechanism. - fail, if the write consistency constrains you have is too strict, and too many nodes are down (rare). Usually the default read and write config are sane. - succeed, if the vclock data is still valid for this key (most of the time)

In case it succeeds, it might be that the network topology was in a split-brain situation. In this case, Riak will solve the issue itself using v-clock data.