I tried netperf / netserver with Dom0 / DomU, but I could''nt get a
correct
NW traffic with xentop. Is there anybody could help me?
At Host1''s Dom0, run netperf -H VMIP
At Host2''s DomU (with VMIP named TTVM), run netserver
At Host1, run xentop this way ->"xentop -n -b -d 1"
At Host2, run xentop this way->"xentop -n -b -d 1"
In Host2''s xentop, I got the DomU''s NETRX and DomU''s
CPU as well as
Dom0''s high load during netperf running. It means DomU accturally
received
the packats from Host1''s Dom0.
TTVM''s NETRX changed this way : 9699596->9709952->9845192
NAME STATE CPU(sec) CPU(%) MEM(k) MEM(%) MAXMEM(k) MAXMEM(%) VCPUS NETS
NETTX(k) NETRX(k) VBDS VBD_OO VBD_RD VBD_WR SSID
Domain-0 -----r 2145 0.7 523008 6.2 no limit n/a 1 0 0 0 0 0 0 0 0
TTVM --b--- 189 0.1 260864 3.1 262144 3.1 1 1 21583488 9699596 1 0 2270
1325 0
NAME STATE CPU(sec) CPU(%) MEM(k) MEM(%) MAXMEM(k) MAXMEM(%) VCPUS NETS
NETTX(k) NETRX(k) VBDS VBD_OO VBD_RD VBD_WR SSID
Domain-0 -----r 2145 8.2 523008 6.2 no limit n/a 1 0 0 0 0 0 0 0 0
TTVM -----r 189 4.8 260864 3.1 262144 3.1 1 1 21583672 9709952 1 0 2270
1325 0
NAME STATE CPU(sec) CPU(%) MEM(k) MEM(%) MAXMEM(k) MAXMEM(%) VCPUS NETS
NETTX(k) NETRX(k) VBDS VBD_OO VBD_RD VBD_WR SSID
Domain-0 -----r 2146 100.5 523008 6.2 no limit n/a 1 0 0 0 0 0 0 0 0
TTVM -----r 189 59.4 260864 3.1 262144 3.1 1 1 21586024 9845192 1 0 2270
1325 0
But packat sender Host1''s CPU changed this way :
0.4%->71.6%->100.5
(because it with 2 vcpus)
I could not get packats sending out from Host1. The NETTX no changes
anymore 0->0->0.
Domain-0 -----r 18938 0.4 523008 6.2 no limit n/a 2 0 0 0 0 0 0 0 0
NAME STATE CPU(sec) CPU(%) MEM(k) MEM(%) MAXMEM(k) MAXMEM(%) VCPUS NETS
NETTX(k) NETRX(k) VBDS VBD_OO VBD_RD VBD_WR SSID
Domain-0 -----r 18938 71.6 523008 6.2 no limit n/a 2 0 0 0 0 0 0 0 0
NAME STATE CPU(sec) CPU(%) MEM(k) MEM(%) MAXMEM(k) MAXMEM(%) VCPUS NETS
NETTX(k) NETRX(k) VBDS VBD_OO VBD_RD VBD_WR SSID
Domain-0 -----r 18939 100.5 523008 6.2 no limit n/a 2 0 0 0 0 0 0 0 0
NAME STATE CPU(sec) CPU(%) MEM(k) MEM(%) MAXMEM(k) MAXMEM(%) VCPUS NETS
NETTX(k) NETRX(k) VBDS VBD_OO VBD_RD VBD_WR SSID
And netperf ran well and give a correct NW bencmark.
I could understand the strange issue, what''s wrong?
Is there any network interface or vif''s configuration perhaps I
missed?
Any help is much appreciated.
L.Wang
_______________________________________________
Xen-users mailing list
Xen-users@lists.xensource.com
http://lists.xensource.com/xen-users
2010/1/6 明石 れい <r-akashi@pb.jp.nec.com>:> In Host2''s xentop, I got the DomU''s NETRX and DomU''s CPU as well as > Dom0''s high load during netperf running. It means DomU accturally received > the packats from Host1''s Dom0.> But packat sender Host1''s CPU changed this way : 0.4%->71.6%->100.5 > (because it with 2 vcpus) > I could not get packats sending out from Host1. The NETTX no changes > anymore 0->0->0.Is Host1 a HVM (fully virtualized) domU? -- Fajar _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
Thank you! Fujar.> >> In Host2''s xentop, I got the DomU''s NETRX and DomU''s CPU as well as >> Dom0''s high load during netperf running. It means DomU accturally received >> the packats from Host1''s Dom0. >> > > > >> But packat sender Host1''s CPU changed this way : 0.4%->71.6%->100.5 >> (because it with 2 vcpus) >> I could not get packats sending out from Host1. The NETTX no changes >> anymore 0->0->0. >> > > Is Host1 a HVM (fully virtualized) domU? > >Host1 and Host2 are para-virtualized Dom0. The TTVM is DomU running at Host2. All of them running at a local network. _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
I''ve not found out the clue yet...
It seems like that the xen-3.4.1''s xentop catch Dom0''s NW I/O
from vif0.0.
Check out the xentop source in xen3.4.1,
/usr/src/xen-3.4.1/tools/xenstat/libxenstat/src/xenstat_linux.c
you will find the code that tells me the xentop catch NW traffic from vif.
---------------------------------------------------------
int ret = fscanf(priv->procnetdev,
"vif%u.%u:%llu%llu%llu%llu%*u%*u%*u%*u"
"%llu%llu%llu%llu%*u%*u%*u%*u\n",
&domid, &net.id,
&net.tbytes, &net.tpackets, &net.terrs,
&net.tdrop,
&net.rbytes, &net.rpackets, &net.rerrs,
&net.rdrop);
---------------------------------------------------------
But in my xen-3.4.1 env, Dom0''s bridge name is eth0, but it does NOT
attached to vif0.0. This is what the brctl showing.
--------------------------------------------------------
[root@host_new scripts]# brctl show
bridge name bridge id STP enabled interfaces
eth0 8000.003013e3e302 no peth0
--------------------------------------------------------
In another host, it running on xen-3.1.0, bridge named xenbr0 was attached
to peth0 as well as vif0.0. So, I got correct NW I/O for Dom0.
---------------------------------------------------------
[root@host_old ~]# brctl show
bridge name bridge id STP enabled interfaces
xenbr0 8000.feffffffffff no peth0
vif0.0
----------------------------------------------------------
I just use the default /etc/xen/xend-config.sxp, there xen bridge was
really set to xen bridge mode, the same as it in xen-3.1.0''s same
config
file.
------------------------------
...
(network-script network-bridge)
...
(vif-script vif-bridge)
...
------------------------------
Here is the difference between xen3.1.0 and xen3.4.1.
-----------------------------
◎host_old(=Xen3.1.0)
##
# To bridge network traffic, like this:
#
# dom0: fake eth0 -> vif0.0 -+
# |
# bridge -> real eth0 -> the network
# |
# domU: fake eth0 -> vifN.0 -+
#
# use
#
# (network-script network-bridge)
◎host_new(=Xen3.4.1)
##
# To bridge network traffic, like this:
#
# dom0: ----------------- bridge -> real eth0 -> the network
# |
# domU: fake eth0 -> vifN.0 -+
#
# use
#
# (network-script network-bridge)
-------------------------------
What''s wrong with me?
Any advice is appriciated.
> Thank you! Fujar.
>
>>
>>
>>> In Host2''s xentop, I got the DomU''s NETRX and
DomU''s CPU as well as
>>> Dom0''s high load during netperf running. It means DomU
accturally received
>>> the packats from Host1''s Dom0.
>>>
>>>
>>
>>
>>> But packat sender Host1''s CPU changed this way :
0.4%->71.6%->100.5
>>> (because it with 2 vcpus)
>>> I could not get packats sending out from Host1. The NETTX no
changes
>>> anymore 0->0->0.
>>>
>>>
>> Is Host1 a HVM (fully virtualized) domU?
>>
>>
>>
> Host1 and Host2 are para-virtualized Dom0.
> The TTVM is DomU running at Host2.
> All of them running at a local network.
>
>
>
_______________________________________________
Xen-users mailing list
Xen-users@lists.xensource.com
http://lists.xensource.com/xen-users
2010/1/15 明石 れい <r-akashi@pb.jp.nec.com>:> you will find the code that tells me the xentop catch NW traffic from vif.Ah, OK, makes sense> But in my xen-3.4.1 env, Dom0''s bridge name is eth0, but it does NOT > attached to vif0.0.correct. That''s the new default setup.> Here is the difference between xen3.1.0 and xen3.4.1. > What''s wrong with me?IIRC, the old setup with vif0.0 was deprecated. If you still need it, you might be able to copy /etc/xen/scripts/network-bridge from Xen 3.1.0 and use it on Xen 3.4.x. I have to ask though, why do you need to get dom0''s stats from xentop? Why not read stats from eth0 (the bridge) directly? -- Fajar _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
Many thanks Fujar, I will try your suggestion later.>> you will find the code that tells me the xentop catch NW traffic from vif. >> > > Ah, OK, makes sense > > >> But in my xen-3.4.1 env, Dom0''s bridge name is eth0, but it does NOT >> attached to vif0.0. >> > > correct. That''s the new default setup. > >But the xentop also read NETTX and NETRX from vif , why not from eth0 bridge in xen-3.4.1? BTW, for DomU, xentop tell me the correct NW traffic value ( because vif is generated at Dom0''s side).>> Here is the difference between xen3.1.0 and xen3.4.1. >> What''s wrong with me? >> > > IIRC, the old setup with vif0.0 was deprecated. If you still need it, > you might be able to copy /etc/xen/scripts/network-bridge from Xen > 3.1.0 and use it on Xen 3.4.x. > > I have to ask though, why do you need to get dom0''s stats from xentop? > Why not read stats from eth0 (the bridge) directly? > >My old test program run well at xen-3.1.0. I used it to monitor DomU and Dom0''s CPU, NW loading at every time ( like this way -> xentop -b -d 1). So I felt it''s easier to find out DomU and Dom0''s CPU loading''s change trend when I put NW load at DomU. xentop is very useful monitoring utility for me, if it gives some harddisk load monitoring, it would be more effective monitor utility. Is there any another utility or xen API better than xentop that can monitor CPU, NW, HDD loading/sec ? _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
2010/1/18 明石 れい <r-akashi@pb.jp.nec.com>:>>> But in my xen-3.4.1 env, Dom0''s bridge name is eth0, but it does NOT >>> attached to vif0.0. >>> >> >> correct. That''s the new default setup. >> >> > > But the xentop also read NETTX and NETRX from vif , why not from eth0 > bridge in xen-3.4.1?Because eth0 in xen-3.4.x is not vif. It''s a bridge. This is different from 3.1.x, where eth0 is basically vif0.0 that got renamed to eth0. And if what you wrote earlier is true, xentop only reads data from vif, not bridge. -- Fajar _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
Hi, Fajar xen-3.1.0''s network-bridge''s copy did generate xenbr0 but no vif0.0 generated at the same time. Anyway, thanks for your explanation about the changes around bridge and vif from xen-3.1.X to xen-3.4.X. It give me some clue to fix the problem I call acrossed.>>>> But in my xen-3.4.1 env, Dom0''s bridge name is eth0, but it does NOT >>>> attached to vif0.0. >>>> >>>> >>> correct. That''s the new default setup. >>> >>> >>> >> But the xentop also read NETTX and NETRX from vif , why not from eth0 >> bridge in xen-3.4.1? >> > > Because eth0 in xen-3.4.x is not vif. It''s a bridge. This is different > from 3.1.x, where eth0 is basically vif0.0 that got renamed to eth0. > And if what you wrote earlier is true, xentop only reads data from > vif, not bridge. > >_______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
2010/1/18 明石 れい <r-akashi@pb.jp.nec.com>:> Hi, Fajar > > xen-3.1.0''s network-bridge''s copy did generate xenbr0 but no vif0.0 > generated > at the same time.vif0.0 should be provided by netloop kernel module. verify thata netloop is loaded, also see the output of "brctl show". You should now see that there''s a peth0 there (which is eth0 renamed), eth0 (which is vif0.0 renamed), and possibly several vif0.x. -- Fajar _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
On Mon, Jan 18, 2010 at 08:59:12PM +0700, Fajar A. Nugraha wrote:> 2010/1/18 ?$BL@@P!!$l$$ <r-akashi@pb.jp.nec.com>: > > Hi, Fajar > > > > xen-3.1.0''s network-bridge''s copy did generate xenbr0 but no vif0.0 > > generated > > at the same time. > > vif0.0 should be provided by netloop kernel module. verify thata > netloop is loaded, also see the output of "brctl show". You should now > see that there''s a peth0 there (which is eth0 renamed), eth0 (which is > vif0.0 renamed) >I think you meant to say "eth0 (which is veth0 renamed)" :) -- Pasi _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
On Mon, Jan 18, 2010 at 9:16 PM, Pasi Kärkkäinen <pasik@iki.fi> wrote:> On Mon, Jan 18, 2010 at 08:59:12PM +0700, Fajar A. Nugraha wrote: >> You should now >> see that there''s a peth0 there (which is eth0 renamed), eth0 (which is >> vif0.0 renamed) >> > > I think you meant to say "eth0 (which is veth0 renamed)" :)Yeah :D _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
Hi, Fajar In my environment, kernel /boot/xen-3.4.1.gz ... module /boot/vmlinuz-2.6.18.8-xen ...>> Hi, Fajar >> >> xen-3.1.0''s network-bridge''s copy did generate xenbr0 but no vif0.0 >> generated >> at the same time. >> > > vif0.0 should be provided by netloop kernel module. verify thata > netloop is loaded, also see the output of "brctl show". You should now > see that there''s a peth0 there (which is eth0 renamed), eth0 (which is > vif0.0 renamed), and possibly several vif0.x. > >netloop was not loaded. ----- modprove netloop FATAL: Module netloop not found. ----- Surely, this kernel module does''nt be included in this kernel. xentop told the incorrect Dom0''s NETTX and NETRX from xen3.4.X (or earlier) when I use the default setup. So the best way is to modify the xentop source to get Dom0''s nw traffics from eth0. Thanks. _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
2010/1/19 明石 れい <r-akashi@pb.jp.nec.com>:> netloop was not loaded. > ----- > modprove netloop > FATAL: Module netloop not found. > ----- > Surely, this kernel module does''nt be included in this kernel.The easiest way is to get this module, possibly by recompiling your kernel. RHEL5''s kernel-xen has it.> So the best way is to modify > the xentop source to get Dom0''s nw traffics from eth0.Possibly, if you know how to modify the source. -- Fajar _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users