Red Hat Bugzilla – Bug 181452
Slow multiple reads from Hitachi 9570V
Last modified: 2012-06-20 09:21:00 EDT
From Bugzilla Helper:
User-Agent: Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:126.96.36.199) Gecko/20060111 Firefox/188.8.131.52
Description of problem:
During performance testing and acceptance testing of a Hitachi Data Systems 9570V issues were found regarding the performance of Redhat enterprise when tests were run to perform multiple reads from the SAN.
Indicitive numbers from Windows and Solaris systems when testing for 1,2,3 and 4 reads using dd average read performance was approx 140MB/sec, 137MB/sec, 130MB/sec, 128MB/sec
When testing on RHEL4 for 1,2,3 and 4 reads performance was approx
130MB/sec, 70MB/sec, 78MB/sec, 85MB/sec
See attachments for config files uses during the testing. Installing a vanila Linux kernel, ie fetching kernel source (2.6.15-2) from kernel.org and compiling using RHEL .config file resulted in performance figures in line with Solaris and Windows.
Version-Release number of selected component (if applicable):
All EL series kernels for RHEL4 released by RedHat
Steps to Reproduce:
1. Installed 2x QLA2340 single port HBA's into PowerEdge 1850 Running RHEL4
2. setup multipathd as per attached multipath.conf file
3. used fdisk on /dev/mapper/1HITACHI_D60091120019 to create single 100GB partition
4. Ran kpartx -a /dev/mapper/1HITACHI_D60091120019 to expose partition
5. mounted partition as /SAN
6. Ran the following to create a set of test files time dd if=/dev/zero of=/SAN/test13 bs=8192k count=250
7. Unmounted /SAN after creating four files
8. remounted /SAN
9. see test-commands.txt for tests ran
Actual Results: Results with EL series kernel
Single Read 130MB/sec
Two reads 70MB/sec
Three reads 78MB/sec
Four reads 85MB/sec
Expected Results: These results were acheived while running 2.6.15-2
Single Read 130MB/sec
Dual Read 142MB/sec
Three Reads 113MB/sec
Four Reads 108MB/sec
Each read test was performed with dd, after each test the /SAN filesystem was unmounted and remounted.
SAN is a Hitachi 9570V
Disk used for test was a 100GB LUSE carved out of a 4D+1P FCAL 5x300GB RAID Group
CTRL1 was the owner controller
Both controllers are connected to identical McData 4500 switches
Server was able to see all four paths from HDS Array.
During testing the array uses the correct owner paths.
Created attachment 124607 [details]
multipath daemon config
Created attachment 124608 [details]
HDS prio_callout script to assign owner path correctly
Created attachment 124609 [details]
multipath -lll output, dmesg content RE qla
Thank you for submitting this issue for consideration in Red Hat Enterprise Linux. The release for which you requested us to review is now End of Life.
Please See https://access.redhat.com/support/policy/updates/errata/
If you would like Red Hat to re-consider your feature request for an active release, please re-open the request via appropriate support channels and provide additional supporting details about the importance of this issue.