replicat set 0 a 4 node cluster 4 node in cluster with data in memory and store data in disk ,100G but only 30G been used
then add a new node into cluster with data in memory and configure disk size 60G what would happen when the new node join the cluster
Each node is independent. You could change the storage-engine one node at a time, as long as you have enough capacity and the different threshold for eviction or stop writes are configured correctly, you would be fine. If I understood your current situation, the new node would then fill up and should use even less than 30 GiB since you are adding a node and the other nodes are using 30GiB. If you have expirable data and an eviction threshold set (high-water-disk-pct), make sure your eviction threshold is set so that you do not cross it on the new node as the disk percent used would be higher than the other nodes.
Thank you very much , actually , i am looking for a way to reduce disk dynamiclly , i hope to reducing one node disk a time , our replication factor is 1 , we can accept one node data unaccessible at a period of time .
i think if i can reduce the disk size once at a time , I ask this question because I find the aerospike process will preallocated the disk when it start , ls -al will find the file size will be the value i had set in the configure file ,even if i don’t write any data to AS. so , i fear to lossing data if i change the file size after the data file has been allocated .
i also want to know why the documentation says ,resize the disk need a cold restart .what risk or error would happen if i use rolling restart .
Look forward to your reply : )
after do some search , I understood what you say , we can reconfigure the file size by cold rolling restart .
thank you very much
Yes, you got it!
This topic was automatically closed 84 days after the last reply. New replies are no longer allowed.