On Thu, Jun 25, 2020 at 06:44:19PM +0000, Dmitry Fomichev wrote:
Will these tests succeed if t/zbd/test-zbd-support script is run against an SMR HDD?
Since zoned HDDs don't support Zone Append, I would expect the i/o to fail.
I think you need to check if this device is an NVMe drive and expect the i/o failure in
the tests below if this is not the case.
More inline...
No, these tests fail for SMR HDD.
We will modify the script to run these tests only for NVMe drives.
-----Original Message-----
From: fio-owner@xxxxxxxxxxxxxxx <fio-owner@xxxxxxxxxxxxxxx> On Behalf
Of Krishna Kanth Reddy
Sent: Thursday, June 25, 2020 1:39 PM
To: axboe@xxxxxxxxx
Cc: fio@xxxxxxxxxxxxxxx; Krishna Kanth Reddy <krish.reddy@xxxxxxxxxxx>;
Ankit Kumar <ankit.kumar@xxxxxxxxxxx>
Subject: [PATCH 4/4] t/zbd: Add support to verify Zone Append command
with libaio, io_uring IO engine tests
Modify the test-zbd-support script to verify the Zone Append command
for NVMe Zoned Namespaces (ZNS) defined in NVM Express TP4053.
Added a new FIO option zone_append.
When zone_append option is enabled, the existing write path will
send Zone Append command with LBA offset as start of the Zone.
Signed-off-by: Ankit Kumar <ankit.kumar@xxxxxxxxxxx>
---
t/zbd/test-zbd-support | 48
++++++++++++++++++++++++++++++++++++++++++++++++
1 file changed, 48 insertions(+)
diff --git a/t/zbd/test-zbd-support b/t/zbd/test-zbd-support
index 4001be3..ddade22 100755
--- a/t/zbd/test-zbd-support
+++ b/t/zbd/test-zbd-support
@@ -801,6 +801,54 @@ test48() {
>> "${logfile}.${test_number}" 2>&1 || return $?
}
+# Zone append to sequential zones, libaio, 1 job, queue depth 1
+test49() {
+ local i size
+
+ size=$((4 * zone_size))
+ run_fio_on_seq --ioengine=libaio --iodepth=1 --rw=write --
zone_append=1 \
+ --bs="$(max $((zone_size / 64)) "$logical_block_size")"\
+ --do_verify=1 --verify=md5 \
+ >>"${logfile}.${test_number}" 2>&1 || return $?
+ check_written $size || return $?
+ check_read $size || return $?
+}
+
+# Random zone append to sequential zones, libaio, 8 jobs, queue depth 64
per job
+test50() {
+ local size
+
+ size=$((4 * zone_size))
+ run_fio_on_seq --ioengine=libaio --iodepth=64 --rw=randwrite --bs=4K \
+ --group_reporting=1 --numjobs=8 --zone_append=1 \
+ >> "${logfile}.${test_number}" 2>&1 || return $?
+ check_written $((size * 8)) || return $?
+}
+
+# Zone append to sequential zones, io_uring, 1 job, queue depth 1
+test51() {
+ local i size
+
+ size=$((4 * zone_size))
+ run_fio_on_seq --ioengine=io_uring --iodepth=1 --rw=write --
zone_append=1 \
+ --bs="$(max $((zone_size / 64)) "$logical_block_size")"\
+ --do_verify=1 --verify=md5 \
+ >>"${logfile}.${test_number}" 2>&1 || return $?
+ check_written $size || return $?
+ check_read $size || return $?
+}
+
+# Random zone append to sequential zones, io_uring, 8 jobs, queue depth
64 per job
+test52() {
+ local size
+
+ size=$((4 * zone_size))
Maybe try some different size? It is the same in all tests.
Sure, we will add different sizes for the tests.
+ run_fio_on_seq --ioengine=io_uring --iodepth=64 --rw=randwrite --
bs=4K \
All tests do 4K i/o, but maybe try to run with a different block size?
It could be a good idea to add a test that will write with bs=ZASL(or MDTS).
Yet another test issuing i/o with bs exceeding the maximum i/o size would
be very useful.
Ok, this input is very helpful. We will implement the tests as per your
review comments.
+ --group_reporting=1 --numjobs=8 --zone_append=1 \
+ >> "${logfile}.${test_number}" 2>&1 || return $?
+ check_written $((size * 8)) || return $?
+}
+
tests=()
dynamic_analyzer=()
reset_all_zones=
--
2.7.4
