One thing to check for is (1) large objects in the cache and (2) multi-range
requests. In this case ATS will read the entire object and filter out that
parts that aren't in the range requests. The standard range acceleration only
works for single range requests. Every object resides on a single disk so if
this is a specific object you would see this heavy read only on that disk.
There is a debug message for this but unfortunately it's "http_trans" which
generates *a lot* of output. If you're willing to do a minor tweak of the
source, look in HttpSM.cc for the line reading "Unable to accelerate range
request" and change the tag from "http_trans" to something more specific such
as "http_trans_range" and then use that as the debug tag.
This may not be the problem, but it is worth taking a look.
On Tuesday, March 17, 2015 4:12 PM, Tommy Lee <[email protected]> wrote:
Hi everybody,
We are using ATS 4.2.1 with raw disks config and volume and hosting configs
enabled. And we're experiencing strange high read rates randomly at one disk.
We use LSI Megaraid configured with RAID 0 ( Write back, Read Ahead, Cached
IO and write cache, strip size 1MB ). My question is: There is someway to see
with debug tags what's is going on ? ATS can be performing some
maintenance tasks that randomly read disk at higher rates ?
IOPs goes up to 450 and read speed reaches (90 MB/s) staying for 30~50
seconds and then goes back to normal.
I would thank if anyone can give some direction where to search for this debug
inside the code.
Best regards,
