5 Essential Tips for Maximizing Your Experience at Nutanix .NEXT for Bloggers
Good to hear that you get it worked out and thank you for thorough description of working conf :) Just one point, move stats from disk to tmpfs - you'll lower your disk io requirements significantly bregs Kirkka
Hi JC, we are using dedicated N/VMware (8k series) cluster for hosting PG's (~80 VM's/PG instances) No compression/dedupe on datastore Probably our requirements for perf are not similar as yours (nevertheless some db's are doing ~1k tps /1M tuples - cache is in help) Our setup shortly is following machine conf itself is "standard" Centos6/7 machines 1-4vCPU and 4-32GB RAM per VM and pg conf is calculated based on vm resources PG data is on separate controller (paravirt) all with LVM and XFS for easier manipulation -1:1 data disk (if we hit disk queue hard, we move to separate controller and 2+ disks) -1:2 wal disk -1:3 xlog disk and -stats on tmpfs increased disk read ahead, swappiness 0 and some more (tuned profile with some tweaks) And some cases using pgbouncer on connection heavy application nodes. So far so good :)
Already have an account? Login
Enter your username or e-mail address. We'll send you an e-mail with instructions to reset your password.
Sorry, we're still checking this file's contents to make sure it's safe to download. Please try again in a few minutes.
Sorry, our virus scanner detected that this file isn't safe to download.