[OpenAFS] replication over slower bandwidth
muayad y
muayad_y@yahoo.com
Thu, 9 Feb 2012 20:05:10 -0800 (PST)
---851882195-1726611758-1328846710=:22040
Content-Type: text/plain; charset=iso-8859-1
Content-Transfer-Encoding: quoted-printable
After changing rx.h and recompile debian source packages I was able to tran=
sfer 4.6 MB in 10 seconds 400KB/s. Not sure if this was mistakenly commente=
d out
=A0diff openafs-1.4.12.1+dfsg/src/rx/rx.h=A0 ./rx.h.org
243c243
<=A0=A0=A0=A0 afs_int32 maxPacketSize;=A0=A0 /* max packet size should be p=
er-connection since */
---
> /*=A0=A0=A0 afs_int32 maxPacketSize;=A0=A0=A0 max packet size should be p=
er-connection since */
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D
with rx.h changed=20
dpkg -i openafs-fileserver_1.4.12.1+dfsg-4_amd64.deb
root@web01:/afs/online#=A0 date;vos release test.vol -verbose -localauth;da=
te
Fri Feb 10 03:28:45 UTC 2012
test.vol=20
=A0=A0=A0 RWrite: 536870952=A0=A0=A0=A0 ROnly: 536870953=20
=A0=A0=A0 number of sites -> 2
=A0=A0=A0=A0=A0=A0 server mastelo partition /vicepa RW Site=20
=A0=A0=A0=A0=A0=A0 server web01 partition /vicepa RO Site=20
This is a complete release of volume 536870952
Cloning RW volume 536870952 to temporary RO... done
Getting status of RW volume 536870952... done
Ending cloning transaction on RW volume 536870952... done
Starting transaction on cloned volume 536870953... done
Updating existing ro volume 536870953 on web01 ...
Starting ForwardMulti from 536870953 to 536870953 on web01 (as of Sun May 3=
1 22:41:01 2420).
Deleting the releaseClone 536870953 ... done
updating VLDB ... done
Released volume test.vol successfully
Fri Feb 10 03:29:18 UTC 2012
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D
With default debian package installation=20
root@web01:/afs/online#=A0 date;vos release test.vol -verbose -localauth;da=
te=20
Fri Feb 10 03:45:01 UTC 2012
test.vol=20
=A0=A0=A0 RWrite: 536870952=A0=A0=A0=A0 ROnly: 536870953=20
=A0=A0=A0 number of sites -> 2
=A0=A0=A0=A0=A0=A0 server mastelo.mdc.gameloft.org partition /vicepa RW Sit=
e=20
=A0=A0=A0=A0=A0=A0 server web01.tac.gameloft.org partition /vicepa RO Site=
=20
This is a complete release of volume 536870952
Cloning RW volume 536870952 to temporary RO... done
Getting status of RW volume 536870952... done
Ending cloning transaction on RW volume 536870952... done
Starting transaction on cloned volume 536870953... done
Updating existing ro volume 536870953 on web01.tac.gameloft.org ...
Starting ForwardMulti from 536870953 to 536870953 on web01.tac.gameloft.org=
(as of Sun May 31 22:51:08 2420).
Deleting the releaseClone 536870953 ... done
updating VLDB ... done
Released volume test.vol successfully
Fri Feb 10 03:45:34 UTC 2012
--- On Wed, 2/8/12, muayad y <muayad_y@yahoo.com> wrote:
From: muayad y <muayad_y@yahoo.com>
Subject: Re: [OpenAFS] replication over slower bandwidth
To: "Neil Davies" <semanticphilosopher@gmail.com>
Cc: openafs-info@openafs.org
Received: Wednesday, February 8, 2012, 2:28 AM
Hi,
=A0 with netperf/UDP tests UDP_STREAM/UDP_RR,=A0 I was able to get better i=
dea about=0Aline speed=A0 with minimal of 400KBPS. Even after using tc and =
with=A0 best scenario happened, where=0Ano retransmit occured, I was able t=
o transfer 4.6 MB in ~35 seconds.=20
=A0=20
There is no disconnect=A0 between the two=A0 nodes,=A0 I have limited=0Atra=
nsfer rate/upload speed/mtu , but with no luck. I do not want to jump=0Aint=
o a conclusion nor make judgment, but how can I fill up the pipe if=0Arxudp=
does not pump much data into it? ~240=A0 bytes/packet
May be someone else can use tc netem to reproduce the problem.
Thanks
---851882195-1726611758-1328846710=:22040
Content-Type: text/html; charset=iso-8859-1
Content-Transfer-Encoding: quoted-printable
<table cellspacing=3D"0" cellpadding=3D"0" border=3D"0" ><tr><td valign=3D"=
top" style=3D"font: inherit;">After changing rx.h and recompile debian sour=
ce packages I was able to transfer 4.6 MB in 10 seconds 400KB/s. Not sure i=
f this was mistakenly commented out<br> diff openafs-1.4.12.1+dfsg/src=
/rx/rx.h ./rx.h.org<br>243c243<br>< afs_in=
t32 maxPacketSize; /* max packet size should be per-connection =
since */<br>---<br>> /* afs_int32 maxPacketSize; =
max packet size should be per-connection since */<br><br>=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D<br>with rx.h changed <br>dpkg -i=
openafs-fileserver_1.4.12.1+dfsg-4_amd64.deb<br>root@web01:/afs/online#&nb=
sp; date;vos release test.vol -verbose -localauth;date<br>Fri Feb 10 03:28:=
45 UTC 2012<br><br>test.vol <br> RWrite: 536870952 &=
nbsp; ROnly: 536870953 <br> number of sites -=
>
2<br> server mastelo partition /vicepa=
RW Site <br> server web01 partition /v=
icepa RO Site <br>This is a complete release of volume 536870952<br>Cloning=
RW volume 536870952 to temporary RO... done<br>Getting status of RW volume=
536870952... done<br>Ending cloning transaction on RW volume 536870952... =
done<br>Starting transaction on cloned volume 536870953... done<br>Updating=
existing ro volume 536870953 on web01 ...<br>Starting ForwardMulti from 53=
6870953 to 536870953 on web01 (as of Sun May 31 22:41:01 2420).<br>Deleting=
the releaseClone 536870953 ... done<br>updating VLDB ... done<br>Released =
volume test.vol successfully<br>Fri Feb 10 03:29:18 UTC 2012<br>=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D<br>With default debian pac=
kage installation <br>root@web01:/afs/online# date;vos release test.v=
ol
-verbose -localauth;date <br>Fri Feb 10 03:45:01 UTC 2012<br><br>test.vol =
<br> RWrite: 536870952 ROnly: 536=
870953 <br> number of sites -> 2<br> =
server mastelo.mdc.gameloft.org partition /vicepa RW Sit=
e <br> server web01.tac.gameloft.org pa=
rtition /vicepa RO Site <br>This is a complete release of volume 536870952<=
br>Cloning RW volume 536870952 to temporary RO... done<br>Getting status of=
RW volume 536870952... done<br>Ending cloning transaction on RW volume 536=
870952... done<br>Starting transaction on cloned volume 536870953... done<b=
r>Updating existing ro volume 536870953 on web01.tac.gameloft.org ...<br>St=
arting ForwardMulti from 536870953 to 536870953 on web01.tac.gameloft.org (=
as of Sun May 31 22:51:08 2420).<br>Deleting the releaseClone 536870953 ...=
done<br>updating VLDB ... done<br>Released volume test.vol
successfully<br>Fri Feb 10 03:45:34 UTC 2012<br><br><br>--- On <b>Wed, 2/8=
/12, muayad y <i><muayad_y@yahoo.com></i></b> wrote:<br><blockquote s=
tyle=3D"border-left: 2px solid rgb(16, 16, 255); margin-left: 5px; padding-=
left: 5px;"><br>From: muayad y <muayad_y@yahoo.com><br>Subject: Re: [=
OpenAFS] replication over slower bandwidth<br>To: "Neil Davies" <semanti=
cphilosopher@gmail.com><br>Cc: openafs-info@openafs.org<br>Received: Wed=
nesday, February 8, 2012, 2:28 AM<br><br><div id=3D"yiv1188550670"><table b=
order=3D"0" cellpadding=3D"0" cellspacing=3D"0"><tbody><tr><td style=3D"fon=
t-family: inherit; font-style: inherit; font-variant: inherit; font-weight:=
inherit; font-size: inherit; line-height: inherit; font-size-adjust: inher=
it; font-stretch: inherit; -x-system-font: none;" valign=3D"top">Hi,<br>&nb=
sp; with netperf/UDP tests UDP_STREAM/UDP_RR, I was able to get bette=
r idea about=0Aline speed with minimal of 400KBPS. Even after using t=
c and with best scenario happened, where=0Ano retransmit occured, I w=
as able to transfer 4.6 MB in ~35 seconds. <br> <br>There is no disco=
nnect between the two nodes, I have limited=0Atransfer ra=
te/upload speed/mtu , but with no luck. I do not want to jump=0Ainto a conc=
lusion nor make judgment, but how can I fill up the pipe if=0Arxudp does no=
t pump much data into it? ~240 bytes/packet<br><br><br>May be someone=
else can use tc netem to reproduce the problem.<br><br>Thanks</td></tr></t=
body></table></div></blockquote></td></tr></table>
---851882195-1726611758-1328846710=:22040--