Quorum issues during rolling upgrade

classic Classic list List threaded Threaded
1 message Options
Reply | Threaded
Open this post in threaded view
|

Quorum issues during rolling upgrade

cass savy
We are upgrading from DSE 3.2 (C*1.2) to 4.5(C*2.0.11) in 6 node cluster with replication factor =3. App uses CL quorum.

Testing the rolling  upgrade in 6 node cluster by upgrading one node at a time. During the upgrade, we have test script that is tyring to simulate prod load to write/read data with quorum. There are no issues when we upgrade the first node. When we bring node 2 down for upgrade , script stopped and ran into quorum issues. App uses java driver 2.1We see this issue with consecutive nodes as well.  We had to restart app multiple times. 

 When we upgraded from 3.1 to 3.2(C* versions underneath were 1.2.6 to 1.2.16 in this case) we didn't have any quorum issues, and also client app was not restarted even once. During this upgrade client was on java driver 1.0.x . We upgraded to 2.1 driver version so that it is compatible with c* upgrade.
Has anyone run into this issue during upgrade from C* 1.2 to 2.0? Quorum should tolerate one node being unavailable in a 6 node cluster,right? Is  there know issues/bugs with driver 2.14?

Please advice.