0

I have a RAID6 array made of 9 SATA3 disks @7200 RPM, 4TB in size. I have them partitioned with Gparted Live as GPT partitions, starting at the first megabyte.

I am concerned about the read speed. The disk wait generated on the array with ±200TPS (reported by iostat 1) goes beyond 20%.

The server is Ubuntu 11.10 ,a file server with 99% of the IO coming from random users downloading random files.

I find this to be abnormally big wait time that forces most nginx processes into D state, allowing for the performance to s*ck.

I would appreciate any ideas where to look at, and i'm thankful in advance for any input.

Hayya
  • 11
  • 2
  • For a given level of performance, the wait time is purely determined by load. You seem to be getting reasonable performance -- 200 TPS. (The heads have to seek.) – David Schwartz Sep 09 '14 at 17:20

1 Answers1

1

Problem solved.

The array is healthy, however the issue was in the config of the NginX server. Sendfile was "off". Upon switching sendfile to "on", the load on the server fell to 1.5-2 (out of 8) , and the TPS of the array is now around ±400. All processes are in "S" state now, where they used to be in "D" prior to switching Sendfile to "on". The peak traffic is to come in a few hours and I will edit this comment should I find anything in this answer is not correct or is being missed.

Hayya
  • 11
  • 2