Alexandre Chapellon | 4 Jun 15:04 2011
Picon

Re: Informations about ggaoed

Hello,

I try to wake this thread up.
I have finally installed my sytem with ggaoed and am very happy with it.
It has really nice features in my opinion, and performances, at least,
fits my needs.
Is there other people using or having already used ggaoed? I'd love to
hear about other experiences... good or bad.

Thanks.

Le jeudi 28 avril 2011 à 13:30 +0200, Alexandre Chapellon a écrit :
> Hello,
> 
> I am in the process of building a virtualisation platform based on Xen
> and AoE.
> Basicly I want 2 Xen hosts witch boot on an AoE SAN (root
> device is /dev/etherd/ex.x). Additionnaly each virtual machine will have
> its filesystem stored on the AoE SAN as well and exported to the dom0s.
> 
> I was more familiar with vblade until now, but i read here and there
> that vblade could missbehave when serving a lot of devices.
> 
> So I now consider using ggaoed, which gives me good results for now.
> Nevertheless I have few questions regarding ggaoed:
> 
> - Am I right to think that the design of ggaoed is more suited to
> serve a lot of devices ( multithread V  forking  model)?
> 
> - Is ggaoed still in active developpement? (last release is now a bit
(Continue reading)

Lars Erik Dangvard Jensen | 25 Jun 17:52 2011

vblade multipath

Hi list

Read several posts about vblade.

I'm trying to figure out if it's a good idea to start 2 or more processes of vblade for multiple NICs, and get
more real throughput? Or go for ggaoed instead (if that's production friendly)?

Thanks.

/Lars

------------------------------------------------------------------------------
All the data continuously generated in your IT infrastructure contains a 
definitive record of customers, application performance, security 
threats, fraudulent activity and more. Splunk takes this data and makes 
sense of it. Business sense. IT sense. Common sense.. 
http://p.sf.net/sfu/splunk-d2d-c1
Paweł Sikora | 25 Jun 18:31 2011
Picon

Re: vblade multipath

On Saturday, June 25, 2011 05:52:23 PM Lars Erik Dangvard Jensen wrote:
> Hi list
> 
> Read several posts about vblade.
> 
> I'm trying to figure out if it's a good idea to start 2 or more processes of vblade for multiple NICs, and get
more real throughput? Or go for ggaoed instead (if that's production friendly)?

be aware! aoe solution is buggy -> https://bugzilla.kernel.org/show_bug.cgi?id=26012

BR,
Paweł.

------------------------------------------------------------------------------
All the data continuously generated in your IT infrastructure contains a 
definitive record of customers, application performance, security 
threats, fraudulent activity and more. Splunk takes this data and makes 
sense of it. Business sense. IT sense. Common sense.. 
http://p.sf.net/sfu/splunk-d2d-c1
_______________________________________________
Aoetools-discuss mailing list
Aoetools-discuss <at> lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/aoetools-discuss
Lars Erik Dangvard Jensen | 25 Jun 21:10 2011

Re: vblade multipath

Den 25/06/2011 kl. 18.31 skrev Paweł Sikora:

> On Saturday, June 25, 2011 05:52:23 PM Lars Erik Dangvard Jensen wrote:
>> Hi list
>> 
>> Read several posts about vblade.
>> 
>> I'm trying to figure out if it's a good idea to start 2 or more processes of vblade for multiple NICs, and get
more real throughput? Or go for ggaoed instead (if that's production friendly)?
> 
> be aware! aoe solution is buggy -> https://bugzilla.kernel.org/show_bug.cgi?id=26012

Ok seems like it's aoe version 47, latest from coraid.com is 77 last time I looked. So upgrading aoe driver
could be a good way to test if that solves the issue.
aoe in upstream kernel always seems to be outdated, but it's pretty easy to compile for custom kernels.

/Lars 
------------------------------------------------------------------------------
All the data continuously generated in your IT infrastructure contains a 
definitive record of customers, application performance, security 
threats, fraudulent activity and more. Splunk takes this data and makes 
sense of it. Business sense. IT sense. Common sense.. 
http://p.sf.net/sfu/splunk-d2d-c1
_______________________________________________
Aoetools-discuss mailing list
Aoetools-discuss <at> lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/aoetools-discuss
Paweł Sikora | 26 Jun 11:04 2011
Picon

Re: vblade multipath

On Saturday, June 25, 2011 09:10:31 PM Lars Erik Dangvard Jensen wrote:
> Den 25/06/2011 kl. 18.31 skrev Paweł Sikora:
> 
> > On Saturday, June 25, 2011 05:52:23 PM Lars Erik Dangvard Jensen wrote:
> >> Hi list
> >> 
> >> Read several posts about vblade.
> >> 
> >> I'm trying to figure out if it's a good idea to start 2 or more processes of vblade for multiple NICs, and
get more real throughput? Or go for ggaoed instead (if that's production friendly)?
> > 
> > be aware! aoe solution is buggy -> https://bugzilla.kernel.org/show_bug.cgi?id=26012
> 
> Ok seems like it's aoe version 47, latest from coraid.com is 77 last time I looked. So upgrading aoe driver
could be a good way to test if that solves the issue.
> aoe in upstream kernel always seems to be outdated, but it's pretty easy to compile for custom kernels.

aoe v77 doesn't compile on the top of the 2.6.39.2 (needs part of the linux BKL removal: commit 2a48fc0ab24241755dc93bfd4f01d68efab47f5a
Author: Arnd Bergmann, block: autoconvert trivial BKL users to private mutex). after manual fix, it
simply doesn't work:

[  353.472719] aoe: AoE v77 initialised.
[  353.473282] aoe: warning: unknown aoe command type 0x6b

------------------------------------------------------------------------------
All of the data generated in your IT infrastructure is seriously valuable.
Why? It contains a definitive record of application performance, security 
threats, fraudulent activity, and more. Splunk takes this data and makes 
sense of it. IT sense. And common sense.
http://p.sf.net/sfu/splunk-d2d-c2
(Continue reading)

Alexandre Chapellon | 30 Jun 17:13 2011
Picon

udev warnings

hello,

After I installed aoe driver from coraid (version 76) I got the following warning messages upon reboot (when udev starts):

udevd-work[608]: kernel-provided name 'discover' and NAME= 'etherd/discover' disagree, please use SYMLINK+= or change the kernel to provide the proper name
udevd-work[605]: kernel-provided name 'err' and NAME= 'etherd/err' disagree, please use SYMLINK+= or change the kernel to provide the proper name
udevd-work[610]: kernel-provided name 'flush' and NAME= 'etherd/flush' disagree, please use SYMLINK+= or change the kernel to provide the proper name
udevd-work[593]: kernel-provided name 'revalidate' and NAME= 'etherd/revalidate' disagree, please use SYMLINK+= or change the kernel to provide the proper name

I'm not an udev expert but found that since version 154 KERNEL and NAME must match.
I have tried to modify  the /etc/udev/rules.d/60-aoe.rules file by changing KERNEL=="stuff" to KERNEL=="etherd/stuff" but I still have the same messages.
I also tried, to set add SYMLINK+="stuff" and it still complains.

Warning messages doesn't fear me by themselves. But what makes me unconfortables is that devices nodes err, interfaces, revalidate and flush are now in /dev/ *AND* /dev/etherd/ (unless SYMLINK+="stuff" is set, in wich case /dev/stuff is a symlink to /dev/etherd/stuff)!
I guess this could lead to some issues.
There is apparently no way to set the kernel name for udev in userspace... does this is part of the kernel code? shouldn't it be updated?

Regards.

P.S: I am using aoe6-76 on debian 6.0 with udev 164 and kernel 2.6.32-5-xen-amd64


--
Attachment (a_chapellon.vcf): text/x-vcard, 403 bytes
------------------------------------------------------------------------------
All of the data generated in your IT infrastructure is seriously valuable.
Why? It contains a definitive record of application performance, security 
threats, fraudulent activity, and more. Splunk takes this data and makes 
sense of it. IT sense. And common sense.
http://p.sf.net/sfu/splunk-d2d-c2
_______________________________________________
Aoetools-discuss mailing list
Aoetools-discuss@...
https://lists.sourceforge.net/lists/listinfo/aoetools-discuss
Lars Erik Dangvard Jensen | 1 Jul 15:45 2011

ggaoed and multiple NICs

Hi list,

Using ggaoed and can't get rates higher than 2x1gbit even if I have 4 NICs in both ends (target and initiator).

I've tried to test 1 NIC only in both ends, and then with 2 and this gives results as expected, but with 3 and 4
NICs active I don't get any higher than with 2 NICs.

Is there a limitation with ggaoed (target) or the aoe-driver (initiator) in regard to send over more than 2 NICs?

Maybe related to a bug of some kind (or a misconfiguration by me)?

ggaoectl stats shows:

# Statistics for device aoestorage1-90-1-600r10
read_cnt: 4910643
read_bytes: 31261639168
read_time: 6218.2
write_cnt: 7579476
write_bytes: 31045533696
write_time: 3104.74
other_cnt: 8654
other_time: 0.19019
io_slots: 3826646
io_runs: 2265710
queue_length: 270742348
queue_stall: 0
queue_over: 3078086
ata_err: 0
proto_err: 0

# Statistics for device aoestorage1-90-2-2000r6
read_cnt: 20044
read_bytes: 174278656
read_time: 3.21704
write_cnt: 0
write_bytes: 0
write_time: 0
other_cnt: 9102
other_time: 0.194888
io_slots: 20044
io_runs: 14996
queue_length: 17729
queue_stall: 0
queue_over: 0
ata_err: 24
proto_err: 0

# Statistics for interface eth5
rx_cnt: 8377191
rx_bytes: 23632250444
rx_runs: 1452244
rx_buffers_full: 0
tx_cnt: 8378961
tx_bytes: 18443336628
tx_runs: 1439317
tx_buffers_full: 0
dropped: 0
ignored: 0
broadcast: 1768

# Statistics for interface eth6
rx_cnt: 2660454
rx_bytes: 5222574120
rx_runs: 556340
rx_buffers_full: 0
tx_cnt: 2661852
tx_bytes: 8605530440
tx_runs: 626131
tx_buffers_full: 0
dropped: 0
ignored: 0
broadcast: 1396

# Statistics for interface eth7
rx_cnt: 1316281
rx_bytes: 3010471988
rx_runs: 317804
rx_buffers_full: 0
tx_cnt: 1317535
tx_bytes: 4096917300
tx_runs: 398872
tx_buffers_full: 0
dropped: 0
ignored: 0
broadcast: 1252

--

-- 
Med venlig hilsen / Best regards
Lars E. D. Jensen
DCmedia Hosting
Blangstedgårdsvej 8,  DK-5220 Odense SØ
https://www.dcmediahosting.com
+45 8888 7899
lars@...
http://www.linkedin.com/in/larsedj

------------------------------------------------------------------------------
All of the data generated in your IT infrastructure is seriously valuable.
Why? It contains a definitive record of application performance, security 
threats, fraudulent activity, and more. Splunk takes this data and makes 
sense of it. IT sense. And common sense.
http://p.sf.net/sfu/splunk-d2d-c2

Gmane