Re: FIO - Client and Server - Suggestion

[Jens Axboe <axboe@kernel.dk>]

On 2014-10-10 07:32, Neto, Antonio Jose Rodrigues wrote:
>
>
> On 10/8/14, 10:52 AM, "Jens Axboe" <axboe@kernel.dk> wrote:
>
>> On 10/08/2014 08:47 AM, Neto, Antonio Jose Rodrigues wrote:
>>>
>>>
>>> On 10/8/14, 10:33 AM, "Jens Axboe" <axboe@kernel.dk> wrote:
>>>
>>>> On 10/08/2014 08:13 AM, Neto, Antonio Jose Rodrigues wrote:
>>>>>
>>>>>
>>>>> On 10/8/14, 12:03 AM, "Jens Axboe" <axboe@kernel.dk> wrote:
>>>>>
>>>>>> On 2014-10-07 21:24, Neto, Antonio Jose Rodrigues wrote:
>>>>>>> Nossa Senhora:fio neto$ ./fio --client 10.61.109.151 --remote-config
>>>>>>> /root/fio.patch/fio/model
>>>>>>> hostname=s1, be=0, 64-bit, os=Linux, arch=x86-64,
>>>>>>> fio=fio-2.1.13-42-g3232,
>>>>>>> flags=1
>>>>>>> <s1> fio: unable to open '/root/fio.patch/fio/model:70?' job file
>>>>>>> client: host=10.61.109.151 disconnected
>>>>>>>
>>>>>>> Any ideas?
>>>>>>
>>>>>> Looks like I just forgot to zero terminate that string. It was never
>>>>>> absolute or relative path, just luck and what was in memory. Try and
>>>>>> pull again, I committed a fix for that.
>>>>>>
>>>>>> --
>>>>>> Jens Axboe
>>>>>>
>>>>>> --
>>>>>
>>>>>
>>>>> Hi Jens,
>>>>>
>>>>> This is neto from Brazil
>>>>>
>>>>> How are you?
>>>>>
>>>>> Seems to me it's working with absolute path now with the latest commit
>>>>> to
>>>>> remote-config branch.
>>>>
>>>> Great, I verified this morning that it was an issue, we'd be looking at
>>>> unitialized/allocated memory without it.
>>>>
>>>>> But, running the workload from my mac (connected to 2 Linux clients) I
>>>>> do
>>>>> not see the progress.
>>>>>
>>>>> Nossa Senhora:fiop neto$ ./fio --client 10.61.109.151 --remote-config
>>>>> /root/fiop/model --client 10.61.109.152 --remote-config
>>>>> /root/fiop/model
>>>>> hostname=s2, be=0, 64-bit, os=Linux, arch=x86-64, fio=fio-2.1.13,
>>>>> flags=1
>>>>> hostname=s1, be=0, 64-bit, os=Linux, arch=x86-64,
>>>>> fio=fio-2.1.13-31-g15e3,
>>>>> flags=1
>>>>> <s2> workload: (g=0): rw=read, <s1> workload: (g=0): rw=read,
>>>>> bs=64K-64K/64K-64K/64K-64K, bs=64K-64K/64K-64K/64K-64K,
>>>>> ioengine=libaio,
>>>>> iodepth=1
>>>>> ioengine=libaio, iodepth=1
>>>>> <s2> ...
>>>>> <s1> ...
>>>>> <s1> Starting <s2> Starting 128 threads
>>>>> 128 threads
>>>>> Jobs: 0 (f=0)
>>>>>
>>>>> Any idea why?
>>>>
>>>> Works for me, just tried it from an OSX client. I notice that you don't
>>>> seem to have updated the 's2' fio version, however. So I'd suggest you
>>>> ensure you are running the same thing on all of them.
>>>>
>>>> --
>>>> Jens Axboe
>>>>
>>>
>>>
>>> Hi Jens,
>>>
>>> This is neto from Brazil
>>>
>>> How are you?
>>>
>>> With one client and one server it works
>>>
>>> Nossa Senhora:fiop neto$ ./fio --client 10.61.109.151 --remote-config
>>> /root/fiop/model
>>> hostname=s1, be=0, 64-bit, os=Linux, arch=x86-64,
>>> fio=fio-2.1.13-31-g15e3,
>>> flags=1
>>> <s1> workload: (g=0): rw=read, bs=64K-64K/64K-64K/64K-64K,
>>> ioengine=libaio, iodepth=1
>>> <s1> ...
>>> <s1> Starting 128 threads
>>> Jobs: 128 (f=2048): [R(128)] [4.4% done] [1770M/0K/0K /s] [27.7K/0/0
>>> iops]
>>> [eta 09m:45s]
>>>
>>>
>>>
>>>
>>>
>>> But with one client and 2 servers it does not work (the progress)
>>>
>>>
>>> Nossa Senhora:fiop neto$ ./fio --client 10.61.109.151 --remote-config
>>> /root/fiop/model --client 10.61.109.152 --remote-config /root/fiop/model
>>> hostname=s2, be=0, 64-bit, os=Linux, arch=x86-64,
>>> fio=fio-2.1.13-31-g15e3,
>>> flags=1
>>> hostname=s1, be=0, 64-bit, os=Linux, arch=x86-64,
>>> fio=fio-2.1.13-31-g15e3,
>>> flags=1
>>> <s2> workload: (g=0): rw=read, <s1> workload: (g=0): rw=read,
>>> bs=64K-64K/64K-64K/64K-64K, bs=64K-64K/64K-64K/64K-64K, ioengine=libaio,
>>> iodepth=1
>>> ioengine=libaio, iodepth=1
>>> <s2> ...
>>> <s1> ...
>>> <s2> Starting <s1> Starting 128 threads128 threads
>>>
>>> Jobs: 0 (f=0)
>>> Jobs: 0 (f=0)
>>
>> Weird, tested two here, running different jobs, and it summed them up
>> fine and reported the ETA line. I will take a look, when time permits.
>>
>> --
>> Jens Axboe
>
>
> Hi Jens,
>
> This is neto from Brazil
>
> How are you?
>
> Just a quick update to help you with the troubleshooting.
>
>  From latest commit ...
>
> When I start the fio (from my mac to use 2 Linux servers)
>
> When the job starts, I do not see anything on the screen only this:
>
> Nossa Senhora:fio neto$ ./fio --client 10.61.109.151 --remote-config
> /root/fio/write --client 10.61.109.152 --remote-config /root/fio/write
> hostname=s2, be=0, 64-bit, os=Linux, arch=x86-64, fio=fio-2.1.13-58-g3441,
> flags=1
> hostname=s1, be=0, 64-bit, os=Linux, arch=x86-64, fio=fio-2.1.13-58-g3441,
> flags=1
> <s2> workload: (g=0): rw=write, <s1> workload: (g=0): rw=write,
> bs=32K-32K/32K-32K/32K-32K, bs=32K-32K/32K-32K/32K-32K, ioengine=libaio,
> iodepth=4
> ioengine=libaio, iodepth=4
> <s2> ...
> <s1> ...
> <s2> Starting <s1> Starting 64 threads
> 64 threads
> Jobs: 0 (f=0)
>
>
>
>
> After 60 seconds.... (on my config file)
>
> I have this:
>
> workload: (groupid=0, jobs=64): err= 0: pid=3644: Fri Oct 10 09:31:50 2014
>    mixed: io=36714MB, bw=1223.2MB/s, iops=39141, runt= 30015msec
>      slat (usec): min=10, max=308, avg=20.57, stdev= 5.24
>      clat (usec): min=265, max=295734, avg=6496.79, stdev=12165.17
>       lat (usec): min=280, max=295748, avg=6517.62, stdev=12165.19
>      clat percentiles (usec):
>       |  1th=[  868],  5th=[ 1336], 10th=[ 1720], 20th=[ 2384], 30th=[
> 3024],
>       | 40th=[ 3696], 50th=[ 4448], 60th=[ 5344], 70th=[ 6496], 80th=[
> 8096],
>       | 90th=[11968], 95th=[15808], 99th=[22912], 100th=[69120],
> 100th=[197632],
>       | 100th=[211968], 100th=[254976]
>      bw (KB  /s): min= 3648, max=58112, per=1.57%, avg=19618.75,
> stdev=4463.44
>      lat (usec) : 500=0.06%, 750=0.45%, 1000=1.38%
>      lat (msec) : 2=12.27%, 4=29.89%, 10=41.98%, 20=12.22%, 50=1.23%
>      lat (msec) : 100=0.06%, 250=0.45%, 500=0.01%
>    cpu          : usr=0.72%, sys=1.15%, ctx=1146612, majf=0, minf=223
>    IO depths    : 1=0.1%, 2=0.1%, 4=100.0%, 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=1174844/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=4
>
> Run status group 0 (all jobs):
>    MIXED: io=36714MB, aggrb=1223.2MB/s, minb=1223.2MB/s, maxb=1223.2MB/s,
> mint=30015msec, maxt=30015msec
> Jobs: 0 (f=0)
>
>
>
> After 60 seconds.... ( I have this)...
>
>
> <s1>  0 (f=0)
> workload: (groupid=0, jobs=64): err= 0: pid=3607: Fri Oct 10 09:32:04 2014
>    mixed: io=60243MB, bw=1338.6MB/s, iops=42833, runt= 45006msec
>      slat (usec): min=10, max=1097, avg=21.47, stdev= 7.11
>      clat (usec): min=256, max=302936, avg=5944.33, stdev=9841.20
>       lat (usec): min=275, max=302957, avg=5966.03, stdev=9841.14
>      clat percentiles (usec):
>       |  1th=[  820],  5th=[ 1240], 10th=[ 1656], 20th=[ 2416], 30th=[
> 3120],
>       | 40th=[ 4048], 50th=[ 4704], 60th=[ 5152], 70th=[ 6048], 80th=[
> 7328],
>       | 90th=[10304], 95th=[14912], 99th=[21888], 100th=[25728],
> 100th=[181248],
>       | 100th=[207872], 100th=[244736]
>      bw (KB  /s): min= 8256, max=53824, per=1.57%, avg=21476.50,
> stdev=4885.09
>      lat (usec) : 500=0.07%, 750=0.59%, 1000=1.81%
>      lat (msec) : 2=11.88%, 4=25.20%, 10=49.90%, 20=8.94%, 50=1.29%
>      lat (msec) : 100=0.04%, 250=0.26%, 500=0.01%
>    cpu          : usr=0.78%, sys=1.29%, ctx=1952463, majf=0, minf=219
>    IO depths    : 1=0.1%, 2=0.1%, 4=100.0%, 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=1927764/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=4
>
> Run status group 0 (all jobs):
>    MIXED: io=60243MB, aggrb=1338.6MB/s, minb=1338.6MB/s, maxb=1338.6MB/s,
> mint=45006msec, maxt=45006msec

It's weird, like there's some clock source issue. What happens if you 
add --eta=always as an option to fio?

-- 
Jens Axboe