Recommended setting for a large cluster for faster node recovery
atomix
✭
Hi,
We are running a 12 node Vertica cluster (AWS D2.8XL node, 240G of RAM and 36 cores per node), with 48T storage in each. We are in the process of recovering a node, however the process seems extremely slow. We stopped all ETL on the cluster, would just like to recover the node as fast as possible an resume regular operations.
What is the recommended MAXCONCURRENCY AND PLANNEDCONCURRENCY for such sized cluster?
When I run the following query, it shows only 1 projection is recovering at a time.SELECT * FROM projection_recoveries WHERE status='running' order by start_time DESC LIMIT 100;
Currently plannedconcurrency is set to AUTO, and maxconcurrency is set 5. Is there a way to force Vertica to recover multiple tables/projections in parallel? When I check on CPU and memory usage, I see that there is hardly any resources are used.
Thank you for any recommendations.
Answers
@atomix setting looks ok.
what's the vertica version select version();
select recoverbytable from vs_global_settings;
For better understanding about recovery progress we would suggest running the below query
select * from recovery_status;
select * from table_recoveries;
Thank you! Did it.