Greetings. I have a RedHat 5.7 64bit virtual server on VMware ESXi 4.1. This server and other Redhat Servers are running very slow. I did some stats collection on ESXi and looks like Linux is holding the disk IO. I am not sure what is causing this behavior.
On Linux I checked the CPU and Mem utilisation which is very low, however I see high disk utilisation which is because of the activity. To copy 30 GB data it takes 7 hours. We are using NOOP disk schedular. This has been recommended by both vmware and redhat. Its not the network because the copy source is on the same virtual infrastructure. We have AMS2500 Storage array. I also sent the perf logs to Hitachi and they did not find any latencies on the storage array side. The LQD is 32.
Hi, have you tried the default IO scheduler? How are you copying your data? What is your IO size? How many IOPS are being performed? How is the disk space offered to the VM by ESXi?
Thanks for the quick response. LINUX is doing 4 KB io. Vmware esxi is doing 128 KB. We are using scp to copy the data. I checked on esxi in real time and i noticed 200 - 350 IOPS. Surprisingly the IO is random. Out of 200 -250 only 40 - 50were rights and rest of them were reads. I am not sure what is causing the reads to be so high. This serve is running oracle but the database was stopped during the transfer ans no aplication was accesing the database. The disk is a vmdk and we are using PVA scsi controller. LVM has been used to create volumes and its with default chunk size.