19 Posts
0
902
R710 / H700 Read Performance Drop
Hi,
a few weeks ago (less than a month) I installed 2x R710 server for internal purposes. Both have the same hardware configuration and running Windows 2019 LTSC. Each server is equipped with a H700 controller and a raid10 consiting of 4x 6TB (non Dell blessed) Seagate HDDs. All firmwares are the latest available.
In the first week the servers running fine. One day I noticed a dramatically READ performance drop. The avarage read speed was arround 1MB/s while the write speed still reached the max performance. Strange, this phenomenom appeared nearly eaqually on both server. Because I've formatted the raid volume with ReFS my first thought was a bug with the filesystem.
So I decided kill the raid and rebuilt it again, now with NTFS. Time goes by and this weekend the same issue appeared. Again nearly at the same time on both server. This time I investigated a little more and checked the file system, raid/hdd error messages an so on. But I did'nt found any indication.
While a copy job was still running on one server, reading data from the volume, I decided to start a consitency chek on the raid volume (the first time!). At the moment I started the process the read speed jumps to (nearly) max again and the issue disappeared. I tested this on the 2nd server and again with exactily the same behaviour.
Needless to say that the consistency checks both finished without any error.
Any ideas?
MopedTobias
19 Posts
0
December 7th, 2019 02:00
I guess I've found the problem.
The problem has something to do with Read Patrol and is reproducible on both servers. The read performance drops appearing while a Read Patrol task is running. The reason a constistency check solve the issue is that a consistency check cancels a running Read Patrol task.
Although I don't know why this happen on this servers my solution is to disable the automatic Read Patrol and perform a consistency check twice a month. That is good enough for the purpose of these servers.