Geek Answers Handbook

Loss of cassandra node

I am trying to figure out how to parameterize my cluster with two nodes in order to have an exact copy if one of them does not work ... using these tools to check this: http://www.ecyrd.com/cassandracalculator/

For the following parameters:

Cluster size: 2 / Replication Factor: 2 / Write Level: All / Read Level: One

he gives me the results:

  • Your messages are consistent
  • You can survive a loss without knots .
  • You really read from 1 node every time .
  • You really write 2 nodes each time .
  • Each node contains 100% of your data.

I agree with all of these results, except for one: "You can survive without losing nodes." I don’t understand why the cluster cannot survive due to the loss of the node, if it dies, “Write: ALL” will write only one node to the left, right? Then, when I return the second node, it should synchronize with the 1st, and accept the record again, right?

If I'm wrong, can someone explain to me why?

EDIT:

An additional configuration will give me the same results, with the exception of the% holding:

Cluster size: 3 / Replication Factor: 2 / Write Level: All / Read Level: One
Cluster size: 3 / Replication Factor: 2 / Write Level: Quorum / Read Level: One

Even if I try to match Reads, not Writes:

Cluster size: 2 / Replication Factor: 2 / Write Level: One / Read Level: Quorum

Every time he tells me that I cannot suffer a loss on any node.

+4
source share
2 answers

This is due to the replication rate.

, , , .

: http://www.datastax.com/docs/1.0/cluster_architecture/replication

+5

, : " ". , node, - , "Write: ALL" node, ? , node, 1-, , ?

, - ?

, . node , QUORUM. 2, , , ALL 2, node , .

(, ). 3 node RF 3, QUORUM. , node .

:

: 2/ : 2/ : / :

QUORUM = RF/2 + 1 = 2, .

, RF QUORUM.

0

More articles:


All Articles