Re: FIO - Client and Server - Suggestion

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On 2014-10-07 19:19, Neto, Antonio Jose Rodrigues wrote:


On 10/7/14, 9:07 PM, "Neto, Antonio Jose Rodrigues"
<Antonio.Jose.Rodrigues.Neto@xxxxxxxxxx> wrote:



On 10/7/14, 6:44 PM, "Neto, Antonio Jose Rodrigues"
<Antonio.Jose.Rodrigues.Neto@xxxxxxxxxx> wrote:





On Oct 7, 2014, at 6:33 PM, Jens Axboe <axboe@xxxxxxxxx> wrote:

On 10/07/2014 04:11 PM, Neto, Antonio Jose Rodrigues wrote:


On 10/7/14, 6:09 PM, "Neto, Antonio Jose Rodrigues"
<Antonio.Jose.Rodrigues.Neto@xxxxxxxxxx> wrote:



On 10/7/14, 6:01 PM, "Jens Axboe" <axboe@xxxxxxxxx> wrote:

On 10/07/2014 03:29 PM, Jens Axboe wrote:
On 10/07/2014 11:38 AM, Neto, Antonio Jose Rodrigues wrote:
Hi All,

This is neto from Brazil

How are you?

One small suggestion:

Running Client and Server architecture on FIO, we need to have all
config
files "locally" to run it.


Nossa Senhora:tools neto$ ls
151 152 fio



Nossa Senhora:tools neto$ ./fio --client 10.61.109.151 151
--client
10.61.109.152 152
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

If we are doing a test in an HPC environment with hundred of
files,
wouldn't be easier to try to point to a single location for the
file
and
access it remotely (we won't need to copy it locally).


For example:

./fio --client 10.61.109.151 --remote-config /root/fio/151
--client
10.61.109.152 --remote-config /root/fio/152


Thoughts?

That's a good idea, would not be hard to insert that step of having
the
remote fio server load a local config file. I'll look into that.

Totally untested, but the below is a start. On the client side,
you'd do:

fio --client=server-hostname --remote-config /some/path/to/file

and then the server should attempt to open that. Needs a bit of
error
handling, but the concept should be there.

--
Jens Axboe


Hi Jens,

This is neto from Brazil

How are you?

I think it is not working.... Please see below:

Nossa Senhora:fio.patch neto$ ./fio --client 10.61.109.151
--remote-config
/root/fio.patch/model
hostname=s1, be=0, 64-bit, os=Linux, arch=x86-64,
fio=fio-2.1.13-31-g15e3,
flags=1

Thank you,


neto

On server side, I've got:

[root@s1 fio.patch]# ./fio --server
fio: server listening on 0.0.0.0,8765
fopen job file: No such file or directory
fopen job file: No such file or directory
fopen job file: No such file or directory
fopen job file: No such file or directory
fopen job file: No such file or directory
fopen job file: No such file or directory
fopen job file: No such file or directory
fopen job file: No such file or directory

Just to be sure, when you use --remote-config /root/fio.patch/model,
then it will open /root/fio.patch/model on the s1 machine running fio
--server.

Ran a quick test here, and it seems to work for me. The fio --server
opens the file at the given path, not the client. If you wanted the
client to open it, you'd just do

fio --client 10.61.109.151 /root/fio.patch/model

and it'd do the opposite - load the file on the client side, and send
it
to the server.

--
Jens Axboe


I will try again tonight but it did not work

My client is on mac and servers are on linux



Hi Jens,

This is neto from Brazil

How are you?

I found what is happening

If I use a non absolute path it works: the remote config should be on the
same directory with fio


Nossa Senhora:fio.patch neto$ ./fio --client 10.61.109.151 --remote-config
conf3



If I specify an absolute path, it does not work

Nossa Senhora:fio.patch neto$ ./fio --client 10.61.109.151 --remote-config
/root/fio.patch/conf3
hostname=s1, be=0, 64-bit, os=Linux, arch=x86-64, fio=fio-2.1.13-31-g15e3,
flags=1


Do you think we could fix it to make sure we can use --remote-config with
an absolute path?

Thanks

neto



Also, another thing that I have figured out is (not sure if this was by
design or not)


Nossa Senhora:fio.patch neto$ ./fio --client 10.61.109.151 --remote-config
model --client 10.61.109.152 --remote-config 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, bs=64K-64K/64K-64K/64K-64K, <s1> workload:
(g=0): rw=read, ioengine=libaio, iodepth=1
bs=64K-64K/64K-64K/64K-64K, <s2> ...
ioengine=libaio, iodepth=1
<s1> ...
<s2> Starting <s1> Starting 128 threads
128 threads
Jobs: 0 (f=0)
Jobs: 0 (f=0)


Jobs: 0 (f=0)
Jobs: 0 (f=0)





Running client and server, we do not have the progress (%) and IOPS and
KB/s and the time....

Is this expected?

Nope, that should work as non client/server. It does for me. You might want to try a full make clean and remake, the patch touches some structures and fio isn't always consistent in rebuilding perfectly (this is something that should be fixed...).

--
Jens Axboe

--
To unsubscribe from this list: send the line "unsubscribe fio" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html




[Index of Archives]     [Linux Kernel]     [Linux SCSI]     [Linux IDE]     [Linux USB Devel]     [Video for Linux]     [Linux Audio Users]     [Yosemite News]     [Linux SCSI]

  Powered by Linux