undefined method `path' for 12 tests, 0 assertions, 0 failures, 0 errors:Test::Unit::TestResult

[akiradeveloper]

When I run dmtest run --suite writeboost -t /Type0/ dmtest emits the following message at last.

undefined method `path' for 12 tests, 0 assertions, 0 failures, 0 errors:Test::Unit::TestResult

I suspect this is because some code of terminating suite calls path method on a object whose string representation is "12 tests ...". Do you have the same problem in your environment?

By the way, I don't know whether this relates to the issue above, the command I run doesn't run all tests but only 12 tests are run.

[simonxuhao]

Hi Akira,

How do you know there is a writeboost testsuite? By looking at the directories in lib/dmtest/tests? writeboost is not listed in README.md.

How to get a list of available testsuites?

[jthornber]

Each suite has a file in dmtest/lib/suites/

On Mon, 11 May 2015 at 09:13 Simon Xu notifications@github.com wrote:

Hi Akira,

How do you know there is a writeboost testsuite? By looking at the
directories in lib/dmtest/tests? writeboost is not listed in README.md.

How to get a list of available testsuites?

—
Reply to this email directly or view it on GitHub
#42 (comment)
.

[akiradeveloper]

I think this problem is reproducible with other suite, isn't it? I don't know why it fails after running 12 tests.

[akiradeveloper]

I will send you PR that includes adding writeboost to README.
I thought that only in-tree targets can be listed there but out-of-tree enhanceio is seen.

[jthornber]

I'm happy to have anything you want in dmtest.

On Mon, 11 May 2015 at 09:36 Akira Hayakawa notifications@github.com
wrote:

I will send you PR that includes adding writeboost to README.
I thought that only in-tree targets can be listed there but out-of-tree
enhanceio is seen.

—
Reply to this email directly or view it on GitHub
#42 (comment)
.

[simonxuhao]

I also hit this issue when I run 'dmtest run --suite wirteboost':

...
...
no_read_cache...test: (g=0): rw=read, bs=4K-4K/4K-4K/4K-4K, ioengine=sync, iodepth=1
fio-2.2.8
Starting 1 process

test: (groupid=0, jobs=1): err= 0: pid=2516: Mon May 11 16:23:34 2015
read : io=16384KB, bw=53368KB/s, iops=13342, runt= 307msec
clat (usec): min=39, max=6584, avg=72.46, stdev=155.39
lat (usec): min=39, max=6585, avg=72.65, stdev=155.40
clat percentiles (usec):
| 1.00th=[ 40], 5.00th=[ 41], 10.00th=[ 42], 20.00th=[ 43],
| 30.00th=[ 44], 40.00th=[ 45], 50.00th=[ 47], 60.00th=[ 49],
| 70.00th=[ 101], 80.00th=[ 110], 90.00th=[ 115], 95.00th=[ 117],
| 99.00th=[ 199], 99.50th=[ 253], 99.90th=[ 860], 99.95th=[ 4128],
| 99.99th=[ 6560]
lat (usec) : 50=61.45%, 100=5.44%, 250=32.59%, 500=0.39%, 1000=0.02%
lat (msec) : 4=0.02%, 10=0.07%
cpu : usr=4.58%, sys=15.03%, ctx=4096, majf=0, minf=29
IO depths : 1=100.0%, 2=0.0%, 4=0.0%, 8=0.0%, 16=0.0%, 32=0.0%, >=64=0.0%
submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
complete : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
issued : total=r=4096/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0
latency : target=0, window=0, percentile=100.00%, depth=1

Run status group 0 (all jobs):
READ: io=16384KB, aggrb=53368KB/s, minb=53368KB/s, maxb=53368KB/s, mint=307msec, maxt=307msec
FAIL
undefined method `path' for 12 tests, 0 assertions, 0 failures, 3 errors:Test::Unit::TestResult

[simonxuhao]

A linux git repo needs to be available at /root/linux-github. This should also be documented.

[simonxuhao]

Other dependencies that should be documented:

fio
bcache-tools
iozone
cryptsetup

[akiradeveloper]

Simon,
Thanks for testing but
the tests no_read_cache test fails because driver code and the test code
is inconsistent. But yes, the same problem reproduces since it should be
running until all tests are done (and then reports the errors)

Btw, I am not sure what commit of driver you are using but,
I want to notice that dm-writeboost is now trying to live as
out-of-tree driver (since I failed to persuade DM maintainers to merge it in).

I have just release v1.0.0 yesterday
https://github.com/akiradeveloper/dm-writeboost

and the driver code works well with develop@akiradeveloper/device-mapper-test-suite
with kernel >=3.10

I will send Joe PR with the lastest test code which contains some fixes
due to tiny interface changes.

Joe,
Could you please remove dm-writeboost's code from your linux-2.6 tree?
And thanks for your supporting for a long time.

On 2015/05/11 17:39, Simon Xu wrote:

I also hit this issue when I run 'dmtest run --suite wirteboost':

...
...
no_read_cache...test: (g=0): rw=read, bs=4K-4K/4K-4K/4K-4K, ioengine=sync, iodepth=1
fio-2.2.8
Starting 1 process

test: (groupid=0, jobs=1): err= 0: pid=2516: Mon May 11 16:23:34 2015
read : io=16384KB, bw=53368KB/s, iops=13342, runt= 307msec
clat (usec): min=39, max=6584, avg=72.46, stdev=155.39
lat (usec): min=39, max=6585, avg=72.65, stdev=155.40
clat percentiles (usec):
| 1.00th=[ 40], 5.00th=[ 41], 10.00th=[ 42], 20.00th=[ 43],
| 30.00th=[ 44], 40.00th=[ 45], 50.00th=[ 47], 60.00th=[ 49],
| 70.00th=[ 101], 80.00th=[ 110], 90.00th=[ 115], 95.00th=[ 117],
| 99.00th=[ 199], 99.50th=[ 253], 99.90th=[ 860], 99.95th=[ 4128],
| 99.99th=[ 6560]
lat (usec) : 50=61.45%, 100=5.44%, 250=32.59%, 500=0.39%, 1000=0.02%
lat (msec) : 4=0.02%, 10=0.07%
cpu : usr=4.58%, sys=15.03%, ctx=4096, majf=0, minf=29
IO depths : 1=100.0%, 2=0.0%, 4=0.0%, 8=0.0%, 16=0.0%, 32=0.0%, >=64=0.0%
submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
complete : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
issued : total=r=4096/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0
latency : target=0, window=0, percentile=100.00%, depth=1

Run status group 0 (all jobs):
READ: io=16384KB, aggrb=53368KB/s, minb=53368KB/s, maxb=53368KB/s, mint=307msec, maxt=307msec
FAIL
undefined method `path' for 12 tests, 0 assertions, 0 failures, 3 errors:Test::Unit::TestResult

---

Reply to this email directly or view it on GitHub:
#42 (comment)

[jthornber]

I'll remove it in a bit. First I want to look at doing the fix ups I
suggested to you. I've just spent a month on dm-cache, and am confident
I'm not going to get much more performance out of it. So I really need a
solution for the streaming write side of caching.

On Mon, 11 May 2015 at 09:53 Akira Hayakawa notifications@github.com
wrote:

Simon,
Thanks for testing but
the tests no_read_cache test fails because driver code and the test code
is inconsistent. But yes, the same problem reproduces since it should be
running until all tests are done (and then reports the errors)

Btw, I am not sure what commit of driver you are using but,
I want to notice that dm-writeboost is now trying to live as
out-of-tree driver (since I failed to persuade DM maintainers to merge it
in).

I have just release v1.0.0 yesterday
https://github.com/akiradeveloper/dm-writeboost

and the driver code works well with develop@akiradeveloper
/device-mapper-test-suite
with kernel >=3.10

I will send Joe PR with the lastest test code which contains some fixes
due to tiny interface changes.

Joe,
Could you please remove dm-writeboost's code from your linux-2.6 tree?
And thanks for your supporting for a long time.

On 2015/05/11 17:39, Simon Xu wrote:

I also hit this issue when I run 'dmtest run --suite wirteboost':

...
...
no_read_cache...test: (g=0): rw=read, bs=4K-4K/4K-4K/4K-4K,
ioengine=sync, iodepth=1
fio-2.2.8
Starting 1 process

test: (groupid=0, jobs=1): err= 0: pid=2516: Mon May 11 16:23:34 2015
read : io=16384KB, bw=53368KB/s, iops=13342, runt= 307msec
clat (usec): min=39, max=6584, avg=72.46, stdev=155.39
lat (usec): min=39, max=6585, avg=72.65, stdev=155.40
clat percentiles (usec):
| 1.00th=[ 40], 5.00th=[ 41], 10.00th=[ 42], 20.00th=[ 43],
| 30.00th=[ 44], 40.00th=[ 45], 50.00th=[ 47], 60.00th=[ 49],
| 70.00th=[ 101], 80.00th=[ 110], 90.00th=[ 115], 95.00th=[ 117],
| 99.00th=[ 199], 99.50th=[ 253], 99.90th=[ 860], 99.95th=[ 4128],
| 99.99th=[ 6560]
lat (usec) : 50=61.45%, 100=5.44%, 250=32.59%, 500=0.39%, 1000=0.02%
lat (msec) : 4=0.02%, 10=0.07%
cpu : usr=4.58%, sys=15.03%, ctx=4096, majf=0, minf=29
IO depths : 1=100.0%, 2=0.0%, 4=0.0%, 8=0.0%, 16=0.0%, 32=0.0%, >=64=0.0%
submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
complete : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
issued : total=r=4096/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0
latency : target=0, window=0, percentile=100.00%, depth=1

Run status group 0 (all jobs):
READ: io=16384KB, aggrb=53368KB/s, minb=53368KB/s, maxb=53368KB/s,
mint=307msec, maxt=307msec
FAIL
undefined method `path' for 12 tests, 0 assertions, 0 failures, 3
errors:Test::Unit::TestResult

---

Reply to this email directly or view it on GitHub:

#42 (comment)

—
Reply to this email directly or view it on GitHub
#42 (comment)
.

[jthornber]

As for registering dependencies on tools, repo trees etc. I'd like to be
able to annotate code with dependencies so we can produce a list of
dependencies for a given subset of tests.

On Mon, 11 May 2015 at 10:23 Joe Thornber joe.thornber@gmail.com wrote:

I'll remove it in a bit. First I want to look at doing the fix ups I
suggested to you. I've just spent a month on dm-cache, and am confident
I'm not going to get much more performance out of it. So I really need a
solution for the streaming write side of caching.

On Mon, 11 May 2015 at 09:53 Akira Hayakawa notifications@github.com
wrote:

Simon,
Thanks for testing but
the tests no_read_cache test fails because driver code and the test code
is inconsistent. But yes, the same problem reproduces since it should be
running until all tests are done (and then reports the errors)

Btw, I am not sure what commit of driver you are using but,
I want to notice that dm-writeboost is now trying to live as
out-of-tree driver (since I failed to persuade DM maintainers to merge it
in).

I have just release v1.0.0 yesterday
https://github.com/akiradeveloper/dm-writeboost

and the driver code works well with develop@akiradeveloper
/device-mapper-test-suite
with kernel >=3.10

I will send Joe PR with the lastest test code which contains some fixes
due to tiny interface changes.

Joe,
Could you please remove dm-writeboost's code from your linux-2.6 tree?
And thanks for your supporting for a long time.

On 2015/05/11 17:39, Simon Xu wrote:

I also hit this issue when I run 'dmtest run --suite wirteboost':

...
...
no_read_cache...test: (g=0): rw=read, bs=4K-4K/4K-4K/4K-4K,
ioengine=sync, iodepth=1
fio-2.2.8
Starting 1 process

test: (groupid=0, jobs=1): err= 0: pid=2516: Mon May 11 16:23:34 2015
read : io=16384KB, bw=53368KB/s, iops=13342, runt= 307msec
clat (usec): min=39, max=6584, avg=72.46, stdev=155.39
lat (usec): min=39, max=6585, avg=72.65, stdev=155.40
clat percentiles (usec):
| 1.00th=[ 40], 5.00th=[ 41], 10.00th=[ 42], 20.00th=[ 43],
| 30.00th=[ 44], 40.00th=[ 45], 50.00th=[ 47], 60.00th=[ 49],
| 70.00th=[ 101], 80.00th=[ 110], 90.00th=[ 115], 95.00th=[ 117],
| 99.00th=[ 199], 99.50th=[ 253], 99.90th=[ 860], 99.95th=[ 4128],
| 99.99th=[ 6560]
lat (usec) : 50=61.45%, 100=5.44%, 250=32.59%, 500=0.39%, 1000=0.02%
lat (msec) : 4=0.02%, 10=0.07%
cpu : usr=4.58%, sys=15.03%, ctx=4096, majf=0, minf=29
IO depths : 1=100.0%, 2=0.0%, 4=0.0%, 8=0.0%, 16=0.0%, 32=0.0%,
=64=0.0%
submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
complete : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%,
=64=0.0%
issued : total=r=4096/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0
latency : target=0, window=0, percentile=100.00%, depth=1

Run status group 0 (all jobs):
READ: io=16384KB, aggrb=53368KB/s, minb=53368KB/s, maxb=53368KB/s,
mint=307msec, maxt=307msec
FAIL
undefined method `path' for 12 tests, 0 assertions, 0 failures, 3
errors:Test::Unit::TestResult

---

Reply to this email directly or view it on GitHub:

#42 (comment)

—
Reply to this email directly or view it on GitHub
#42 (comment)
.

[akiradeveloper]

Good news to me.

But, I decided not to fix the issues you suggested as the condition
for upsteaming. Since I am confident the 4KB splitting and payload
copyings are inavoidable, I retracted from upstreaming.

As a short summary. For v1.0.0

• I removed Type0, Type1 distinguishment. Type1 is removed so as to simplify
  the code and the document.
• Fix a wq issue to avoid false positive lockdep warning.
  http://www.redhat.com/archives/dm-devel/2013-September/msg00081.html
• Changed message names a bit.
• Start to support 3.10 and later to run in major distributions including RHEL7.
  In 3.14, Kent's bi_iter changed the bio structure. In dm-writeboost,
  the difference between before and after 3.14 is only there.

If you think of streaming write caching,
I want you do it based on writeboost as this researcher also built his
theory based on writeboost (http://embedded.uos.ac.kr/~ysoh/DM-SRC-IBM.pdf)

On 2015/05/11 18:24, Joe Thornber wrote:

I'll remove it in a bit. First I want to look at doing the fix ups I
suggested to you. I've just spent a month on dm-cache, and am confident
I'm not going to get much more performance out of it. So I really need a
solution for the streaming write side of caching.

On Mon, 11 May 2015 at 09:53 Akira Hayakawa notifications@github.com
wrote:

Simon,
Thanks for testing but
the tests no_read_cache test fails because driver code and the test code
is inconsistent. But yes, the same problem reproduces since it should be
running until all tests are done (and then reports the errors)

Btw, I am not sure what commit of driver you are using but,
I want to notice that dm-writeboost is now trying to live as
out-of-tree driver (since I failed to persuade DM maintainers to merge it
in).

I have just release v1.0.0 yesterday
https://github.com/akiradeveloper/dm-writeboost

and the driver code works well with develop@akiradeveloper
/device-mapper-test-suite
with kernel >=3.10

I will send Joe PR with the lastest test code which contains some fixes
due to tiny interface changes.

Joe,
Could you please remove dm-writeboost's code from your linux-2.6 tree?
And thanks for your supporting for a long time.

On 2015/05/11 17:39, Simon Xu wrote:

I also hit this issue when I run 'dmtest run --suite wirteboost':

...
...
no_read_cache...test: (g=0): rw=read, bs=4K-4K/4K-4K/4K-4K,
ioengine=sync, iodepth=1
fio-2.2.8
Starting 1 process

test: (groupid=0, jobs=1): err= 0: pid=2516: Mon May 11 16:23:34 2015
read : io=16384KB, bw=53368KB/s, iops=13342, runt= 307msec
clat (usec): min=39, max=6584, avg=72.46, stdev=155.39
lat (usec): min=39, max=6585, avg=72.65, stdev=155.40
clat percentiles (usec):
| 1.00th=[ 40], 5.00th=[ 41], 10.00th=[ 42], 20.00th=[ 43],
| 30.00th=[ 44], 40.00th=[ 45], 50.00th=[ 47], 60.00th=[ 49],
| 70.00th=[ 101], 80.00th=[ 110], 90.00th=[ 115], 95.00th=[ 117],
| 99.00th=[ 199], 99.50th=[ 253], 99.90th=[ 860], 99.95th=[ 4128],
| 99.99th=[ 6560]
lat (usec) : 50=61.45%, 100=5.44%, 250=32.59%, 500=0.39%, 1000=0.02%
lat (msec) : 4=0.02%, 10=0.07%
cpu : usr=4.58%, sys=15.03%, ctx=4096, majf=0, minf=29
IO depths : 1=100.0%, 2=0.0%, 4=0.0%, 8=0.0%, 16=0.0%, 32=0.0%, >=64=0.0%
submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
complete : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
issued : total=r=4096/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0
latency : target=0, window=0, percentile=100.00%, depth=1

Run status group 0 (all jobs):
READ: io=16384KB, aggrb=53368KB/s, minb=53368KB/s, maxb=53368KB/s,
mint=307msec, maxt=307msec
FAIL
undefined method `path' for 12 tests, 0 assertions, 0 failures, 3
errors:Test::Unit::TestResult

---

Reply to this email directly or view it on GitHub:

#42 (comment)

—
Reply to this email directly or view it on GitHub
#42 (comment)
.

---

Reply to this email directly or view it on GitHub:
#42 (comment)

[jthornber]

ok, let me experiment a bit, then we can have a discussion when I'm more
informed.

On Mon, 11 May 2015 at 10:46 Akira Hayakawa notifications@github.com
wrote:

Good news to me.

But, I decided not to fix the issues you suggested as the condition
for upsteaming. Since I am confident the 4KB splitting and payload
copyings are inavoidable, I retracted from upstreaming.

As a short summary. For v1.0.0

• I removed Type0, Type1 distinguishment. Type1 is removed so as to
  simplify
  the code and the document.
• Fix a wq issue to avoid false positive lockdep warning.
  http://www.redhat.com/archives/dm-devel/2013-September/msg00081.html
• Changed message names a bit.
• Start to support 3.10 and later to run in major distributions including
  RHEL7.
  In 3.14, Kent's bi_iter changed the bio structure. In dm-writeboost,
  the difference between before and after 3.14 is only there.

If you think of streaming write caching,
I want you do it based on writeboost as this researcher also built his
theory based on writeboost (http://embedded.uos.ac.kr/~ysoh/DM-SRC-IBM.pdf
)

On 2015/05/11 18:24, Joe Thornber wrote:

I'll remove it in a bit. First I want to look at doing the fix ups I
suggested to you. I've just spent a month on dm-cache, and am confident
I'm not going to get much more performance out of it. So I really need a
solution for the streaming write side of caching.

On Mon, 11 May 2015 at 09:53 Akira Hayakawa notifications@github.com
wrote:

Simon,
Thanks for testing but
the tests no_read_cache test fails because driver code and the test code
is inconsistent. But yes, the same problem reproduces since it should be
running until all tests are done (and then reports the errors)

Btw, I am not sure what commit of driver you are using but,
I want to notice that dm-writeboost is now trying to live as
out-of-tree driver (since I failed to persuade DM maintainers to merge
it
in).

I have just release v1.0.0 yesterday
https://github.com/akiradeveloper/dm-writeboost

and the driver code works well with develop@akiradeveloper
/device-mapper-test-suite
with kernel >=3.10

I will send Joe PR with the lastest test code which contains some fixes
due to tiny interface changes.

Joe,
Could you please remove dm-writeboost's code from your linux-2.6 tree?
And thanks for your supporting for a long time.

On 2015/05/11 17:39, Simon Xu wrote:

I also hit this issue when I run 'dmtest run --suite wirteboost':

...
...
no_read_cache...test: (g=0): rw=read, bs=4K-4K/4K-4K/4K-4K,
ioengine=sync, iodepth=1
fio-2.2.8
Starting 1 process

test: (groupid=0, jobs=1): err= 0: pid=2516: Mon May 11 16:23:34 2015
read : io=16384KB, bw=53368KB/s, iops=13342, runt= 307msec
clat (usec): min=39, max=6584, avg=72.46, stdev=155.39
lat (usec): min=39, max=6585, avg=72.65, stdev=155.40
clat percentiles (usec):
| 1.00th=[ 40], 5.00th=[ 41], 10.00th=[ 42], 20.00th=[ 43],
| 30.00th=[ 44], 40.00th=[ 45], 50.00th=[ 47], 60.00th=[ 49],
| 70.00th=[ 101], 80.00th=[ 110], 90.00th=[ 115], 95.00th=[ 117],
| 99.00th=[ 199], 99.50th=[ 253], 99.90th=[ 860], 99.95th=[ 4128],
| 99.99th=[ 6560]
lat (usec) : 50=61.45%, 100=5.44%, 250=32.59%, 500=0.39%, 1000=0.02%
lat (msec) : 4=0.02%, 10=0.07%
cpu : usr=4.58%, sys=15.03%, ctx=4096, majf=0, minf=29
IO depths : 1=100.0%, 2=0.0%, 4=0.0%, 8=0.0%, 16=0.0%, 32=0.0%,
=64=0.0%
submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
complete : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%,
=64=0.0%
issued : total=r=4096/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0
latency : target=0, window=0, percentile=100.00%, depth=1

Run status group 0 (all jobs):
READ: io=16384KB, aggrb=53368KB/s, minb=53368KB/s, maxb=53368KB/s,
mint=307msec, maxt=307msec
FAIL
undefined method `path' for 12 tests, 0 assertions, 0 failures, 3
errors:Test::Unit::TestResult

---

Reply to this email directly or view it on GitHub:

#42 (comment)

—
Reply to this email directly or view it on GitHub
<
#42 (comment)

.

---

Reply to this email directly or view it on GitHub:

#42 (comment)

—
Reply to this email directly or view it on GitHub
#42 (comment)
.

[akiradeveloper]

OK.

sync_writes test doesn't work now with the lastest fio and
I suspect regression. But other tests works.

I am looking forward to our discussion.

On 2015/05/11 18:54, Joe Thornber wrote:

ok, let me experiment a bit, then we can have a discussion when I'm more
informed.

On Mon, 11 May 2015 at 10:46 Akira Hayakawa notifications@github.com
wrote:

Good news to me.

But, I decided not to fix the issues you suggested as the condition
for upsteaming. Since I am confident the 4KB splitting and payload
copyings are inavoidable, I retracted from upstreaming.

As a short summary. For v1.0.0

• I removed Type0, Type1 distinguishment. Type1 is removed so as to
  simplify
  the code and the document.
• Fix a wq issue to avoid false positive lockdep warning.
  http://www.redhat.com/archives/dm-devel/2013-September/msg00081.html
• Changed message names a bit.
• Start to support 3.10 and later to run in major distributions including
  RHEL7.
  In 3.14, Kent's bi_iter changed the bio structure. In dm-writeboost,
  the difference between before and after 3.14 is only there.

If you think of streaming write caching,
I want you do it based on writeboost as this researcher also built his
theory based on writeboost (http://embedded.uos.ac.kr/~ysoh/DM-SRC-IBM.pdf
)

On 2015/05/11 18:24, Joe Thornber wrote:

I'll remove it in a bit. First I want to look at doing the fix ups I
suggested to you. I've just spent a month on dm-cache, and am confident
I'm not going to get much more performance out of it. So I really need a
solution for the streaming write side of caching.

On Mon, 11 May 2015 at 09:53 Akira Hayakawa notifications@github.com
wrote:

Simon,
Thanks for testing but
the tests no_read_cache test fails because driver code and the test code
is inconsistent. But yes, the same problem reproduces since it should be
running until all tests are done (and then reports the errors)

Btw, I am not sure what commit of driver you are using but,
I want to notice that dm-writeboost is now trying to live as
out-of-tree driver (since I failed to persuade DM maintainers to merge
it
in).

I have just release v1.0.0 yesterday
https://github.com/akiradeveloper/dm-writeboost

and the driver code works well with develop@akiradeveloper
/device-mapper-test-suite
with kernel >=3.10

I will send Joe PR with the lastest test code which contains some fixes
due to tiny interface changes.

Joe,
Could you please remove dm-writeboost's code from your linux-2.6 tree?
And thanks for your supporting for a long time.

On 2015/05/11 17:39, Simon Xu wrote:

I also hit this issue when I run 'dmtest run --suite wirteboost':

...
...
no_read_cache...test: (g=0): rw=read, bs=4K-4K/4K-4K/4K-4K,
ioengine=sync, iodepth=1
fio-2.2.8
Starting 1 process

test: (groupid=0, jobs=1): err= 0: pid=2516: Mon May 11 16:23:34 2015
read : io=16384KB, bw=53368KB/s, iops=13342, runt= 307msec
clat (usec): min=39, max=6584, avg=72.46, stdev=155.39
lat (usec): min=39, max=6585, avg=72.65, stdev=155.40
clat percentiles (usec):
| 1.00th=[ 40], 5.00th=[ 41], 10.00th=[ 42], 20.00th=[ 43],
| 30.00th=[ 44], 40.00th=[ 45], 50.00th=[ 47], 60.00th=[ 49],
| 70.00th=[ 101], 80.00th=[ 110], 90.00th=[ 115], 95.00th=[ 117],
| 99.00th=[ 199], 99.50th=[ 253], 99.90th=[ 860], 99.95th=[ 4128],
| 99.99th=[ 6560]
lat (usec) : 50=61.45%, 100=5.44%, 250=32.59%, 500=0.39%, 1000=0.02%
lat (msec) : 4=0.02%, 10=0.07%
cpu : usr=4.58%, sys=15.03%, ctx=4096, majf=0, minf=29
IO depths : 1=100.0%, 2=0.0%, 4=0.0%, 8=0.0%, 16=0.0%, 32=0.0%,
=64=0.0%
submit : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%, >=64=0.0%
complete : 0=0.0%, 4=100.0%, 8=0.0%, 16=0.0%, 32=0.0%, 64=0.0%,
=64=0.0%
issued : total=r=4096/w=0/d=0, short=r=0/w=0/d=0, drop=r=0/w=0/d=0
latency : target=0, window=0, percentile=100.00%, depth=1

Run status group 0 (all jobs):
READ: io=16384KB, aggrb=53368KB/s, minb=53368KB/s, maxb=53368KB/s,
mint=307msec, maxt=307msec
FAIL
undefined method `path' for 12 tests, 0 assertions, 0 failures, 3
errors:Test::Unit::TestResult

---

Reply to this email directly or view it on GitHub:

#42 (comment)

—
Reply to this email directly or view it on GitHub
<
#42 (comment)

.

---

Reply to this email directly or view it on GitHub:

#42 (comment)

—
Reply to this email directly or view it on GitHub
#42 (comment)
.

---

Reply to this email directly or view it on GitHub:
#42 (comment)