summaryrefslogtreecommitdiff
path: root/test/README.orig
diff options
context:
space:
mode:
Diffstat (limited to 'test/README.orig')
-rw-r--r--test/README.orig93
1 files changed, 93 insertions, 0 deletions
diff --git a/test/README.orig b/test/README.orig
new file mode 100644
index 0000000..af54c3c
--- /dev/null
+++ b/test/README.orig
@@ -0,0 +1,93 @@
+From the original test/README:
+---------------------------------------------------------------------------
+This directory contains regression suite.
+
+I would appreciate if developer will run it at least once after
+modifications done before commit or mailing list submit.
+
+./regression.sh script expects next binaries and data files exists
+in current directory:
+
+ - regression.dat
+ - iscsiadm
+ - bonnie++ (source: http://www.open-iscsi.org/bits/bonnie++.tar.gz)
+ - disktest (source: http://www.open-iscsi.org/bits/disktest.tar.gz)
+
+Thanks!
+Dmitry
+---------------------------------------------------------------------------
+Call:
+
+> # ./regression.sh -f/--format <device>
+
+to run "mkfs" on the device and exit, or
+
+> # ./regression.sh <targetname> <ipnumber#> <device> [test#[:#]] [bsize]
+
+Where:
+ <targetname> ?
+ <ipnumber#> ?
+ <device> the device on which to test (e.g. /dev/sd?)
+ test#[:#] test(s) to run, i.e. single or range (default: all)
+ bsize disktest block size (default: a range of 10 sizes)
+ special value "bonnie" => skip disktest
+
+And env var "SKIP_WARNING" skips a big warning about writing on the device?
+
+---------------------------------------------------------------------------
+
+The problem is that these tests have not been run for a while, and,
+additionally, disktest seems to be extinct and unfindable.
+
+I plan to get these tests working, and consider these the steps I plan
+to take:
+
+* understand the current tests
+ - particularly, what disktest and bonnie++ are doing
+
+* try to replace disktest and/or bonnie++, if needed
+ - there are a lot of good disk test packages these days
+
+* replace the shell code with PyTest code, to make it
+ easier to use
+
+Lee Duncan -- 2/13/2020
+
+Analysis of disktest usage:
+
+ options used: fio option for this?
+ ======================================= ==================================
+ --name=test (or whatever)
+ -T2 -- run 2 seconds --runtime=2
+ -K8 -- run with 8 threads --numjobs=8
+ -B<bs> -- set block xfer size --blocksize=<bs> (default 4k?)
+ -ID -- use direct IO --direct=1
+ <dev> -- use device --filename=<dev>
+
+ in read mode:
+ -r -- read mode --readwrite=randread
+
+ in write mode:
+ -w -- write mode --readwrite=randwrite
+ -E 16 -- compare 16 bytes --verify=md5? (lots of options)
+
+It looks like the "fio" program may address these needs?
+
+e.g. running
+
+with file "test.fio":
+> [test]
+> rw=randread
+> bs=8k
+> filename=/dev/sdb
+> direct=1
+> numjobs=4
+> runtime=60s
+
+run:
+
+> # fio test.fio
+
+The output is interactive? But results are ridiculously verbose,
+but include a nice summary line or two that could be used as a
+go/no-go?