[MDGeist]

Hole:2 ftp://mir1.ovh.net etch/main bsdutils 1:2.12r-19etch1 [68,5kB]
Hole:3 ftp://mir1.ovh.net etch/main libc6-dev 2.3.6.ds1-13etch5 [2717kB]
2% [3 libc6-dev 539088/2717kB 19%]                              15,1kB/s 39m17s

Downstream almost dead... Below DSL LIGHT niveau!

[buzzz]

Rescue system works very good for me.
iscsi works ok most of the time (sometimes console is not reactive but not that often).
Downstream is fast (sometimes close to 100 mbps).
Upstream seems to be crippled (limited? / I am on SLA) and I get like max 5 mbps (only tried 1 connection @ about 2 am).
Suggestion: What about an OPTION for 100mbps up / 100 mbps down with reasonable traffic limit per month?

[F4RR3LL]

My RPS works since a couple of days without any errors.
Low RAM usage.
Working with SSH is faster but not 100% ok.
Rest works fine.

[gunnarh]

Today booting into rescue-Mode works as expected (got Password by email). Seems to be fixed.

[gunnarh]

When booting into rescue-system still no password-email is sent. only the "your system is up and pingable again" mail is delivered. So rescue-mode is unusable without getting the password.

[gunnarh]

my RPS with Debian, Apache2 and Gallery2 on it sometimes hangs for ~20sec when viewing (small ~150kb JPEG) Pictures. top shows IO/Wait ~100% for a long time.

This does not occour every time, but every few Minutes.

[MDGeist]

not sure how good a benchmark this is but :

 [Rate 200,8/  0,5 KB] [Port: 6892] [U 78/0]

 load average: 1.24, 1.21, 1.34
Tasks:  66 total,   2 running,  64 sleeping,   0 stopped,   0 zombie
Cpu(s):  0.0%us,  0.0%sy,  0.0%ni,  0.0%id, 99.7%wa,  0.0%hi,  0.3%si,  0.0%st

Rtorrent with 78 connections puts the wait time extremely high again (95-99,7), throttling the output i think...
guess it would be a bad idea to run an imageboard on it.
as for many connections at once (like a bigger forum) hmmm, could be troublesome, but cant test it...

[MDGeist]

wasnt it called half duplex? uhm... the one where its only 100 max (so 50-50 in both) and the other one 100-100 ...

edit : looked at wiki

half doesnt make much sense... so i guess its 100mbit in both directions at the same time

edit 2 :
load average: 1.12, 1.17, 0.94
Tasks: 59 total, 2 running, 57 sleeping, 0 stopped, 0 zombie
Cpu(s): 1.3%us, 0.3%sy, 0.0%ni, 0.0%id, 96.0%wa, 0.3%hi, 2.0%si, 0.0%st

[oles@ovh.net]

MDGeist a écrit:
>
> btw. is connection to rps half or full duplex 100mbit ?

I don't know what is the half duplex 100Mbps. does it work ?
someone proposes it ?

[MDGeist]

btw. is connection to rps half or full duplex 100mbit ?

[oles@ovh.net]

MDGeist a écrit:
>
> r10960:/home/rape# time cp KNOPPIX_V5.1.0DVD-2006-12-30-DE.iso
> /dev/null

normal. it means my QoS works. you have a RPS. it means
dedicated CPU/RAM/Network. But the disk is not dedicated.
if you need make lot of write/read (more that web needs)
you will need dedicated server and not RPS.

[MDGeist]

r10960:/home/rape# time cp KNOPPIX_V5.1.0DVD-2006-12-30-DE.iso /dev/null

real 6m25.482s
user 0m0.064s
sys 0m0.936s

r10960:/home/rape# ls -lh
insgesamt 669M
-rw-r--r-- 1 root root 668M 2008-03-26 14:03 KNOPPIX_V5.1.0DVD-2006-12-30-DE.iso
-rw-r--r-- 1 root root 1,9K 2008-03-23 18:31 rape.txt
r10960:/home/rape#



cp command is the worst atm, right ?

and did an upload test
wait is much lower than downloading...
less load, too !

[MDGeist]

im totally raping my rps again...

load beyond 3
wait above 85% !

feels sloooooow ^^

[MDGeist]

Uploading to ISCSI?

will test pure uploading soon... but i think its confusing...

traffic for me is traffic for my rps
nothing intern with i-scsi...

[pendulum]

@traffic manager
its always both up+down ?
im only downloading but manager says im also uploading...

Uploading to ISCSI?

[schwarzlicht]

HI,

Now I will order

Daniel

Good work, it's up (very fast!) - thanks.

[MDGeist]

-/+ buffers/cache:  37  428
load average: 0.06, 0.33, 0.57
Cpu(s):  0.3%us,  0.3%sy,  0.0%ni, 85.0%id, 14.0%wa,  0.0%hi,  0.3%si,  0.0%st

Services running: nginx (1 master, 1 worker), php (1 master, 5 worker), postgresql, munin (master+node), postfix
Applications: rtorrent (seeding debian to a lot of peers), screen, top

=> good machine so far

OT : kannste mal beim outgoing connections thread vorbei schaun? schaffe mit rtorrent nich mehr wie ~30 outgoing cons / torrent

@ rps
it seems as if the 100% waits are now just normal "hangs" ... sytem doesnt update/responds for a short time (<5sek) after that normal again

edit
@ load :
load average: 1.89, 2.51, 2.33

just running wget downloading unix isos

edit 2:
@traffic manager
its always both up+down ?
im only downloading but manager says im also uploading...

[pendulum]

-/+ buffers/cache:  37  428
load average: 0.06, 0.33, 0.57
Cpu(s):  0.3%us,  0.3%sy,  0.0%ni, 85.0%id, 14.0%wa,  0.0%hi,  0.3%si,  0.0%st

Services running: nginx (1 master, 1 worker), php (1 master, 5 worker), postgresql, munin (master+node), postfix
Applications: rtorrent (seeding debian to a lot of peers), screen, top

=> good machine so far

[MDGeist]

hab bisher noch gar nichts drauf außer nen deb4 , upgraded /updated via offiziellem sid repository und rtorrent...

benutze atm grad wget zum dowloadtesten

edit :

concerning new rps :


* higher loads while downloading
* less 100% waits
* system feels slow while heavy IO load, although no 100% wait and free ram/cpu, it doesnt respond fast !

[F4RR3LL]

-/+ buffers/cache:      55148     419436

wow thats realy cool

@MDGeist welche Dienste hast Du installiert bei der geringen 15 MB Speicherbelegung

[schwarzlicht]

HI,

I'm on 194 (BetaTester), no problems since 11 days.
No problems with outgoing Mails (using rps699.ovh.net as MX1 for my domain)

Now I will order

Daniel

[MDGeist]

I'm suprised too. It's not enought ! 15Mb ?? What is your RPS ?

hes quoting my rps again oles, so 10960 again ^^

15mb used when connecting back to shh
pretty low, thats fine isnt it?

[MDGeist]

Downloading at 10mb/s again

waittimes 40-60%

deleting 4gig takes more than a minute
and puts waittimes up to 100% again

load is higher than before the change i think, but i think speed increased too, no sudden high drops due to wait time yet

load right now ~2
when deleting 4gig ~3,3+

[oles@ovh.net]

pendulum a écrit:
>
> MDGeist;18910 Wrote:
>> -/+ buffers/cache: 15028 459556
>> -/+ buffers/cache: 27964 446620
>
> Those numbers are the interesting part MDGeist
>
> 15mb => 28mb used
>
> perfectly fine. (15mb ram used? wow )

I'm suprised too. It's not enought ! 15Mb ?? What is your RPS ?

[oles@ovh.net]

MDGeist a écrit:
>
> Mem: 474584 42716 431868 0 10868
> 16820
> -/+ buffers/cache: 15028 459556
^^^^^^ 459Mb cache
^^^^^ 15Mb used

Linux likes to cache. It's normal.

[oles@ovh.net]

> it was r10960

ahh

[mathias]

...
Just some words about the RPS: the intention with RPS as replacement for Kimsufi is nice. I have experienced a couple of time outs, working at console sometimes impossible, it takes some time till commands will be executed. top about 1 minute, working at console after login 2-3min...
Example these commands:...

Kurze Rückfrage: Wurde von jemandem die Info rausgegeben, die RPS den Isgenug ablösen würden? Das ist nämlich nicht der Fall und hoffentlich führte eine Fehlinformation nicht zu einem allgemeinen Missverständnis.

Mathias

[MDGeist]

oh... whoops, thought it was the free part...

free : 431868
free : 5892

[pendulum]

-/+ buffers/cache: 15028 459556
-/+ buffers/cache: 27964 446620

Those numbers are the interesting part MDGeist

15mb => 28mb used

perfectly fine. (15mb ram used? wow )

[MDGeist]

Mem: 474584 42716 431868 0 10868 16820
-/+ buffers/cache: 15028 459556
Swap: 0 0 0
r10960:/home/rape# screen
[detached]
fr10960:/home/rape# free
total used free shared buffers cached
Mem: 474584 468692 5892 0 4684 436044
-/+ buffers/cache: 27964 446620
Swap: 0 0 0


downloading with 10mb/s drops my ram down 450mb ? wow...

[MDGeist]

MDGeist a écrit:
>
> Feedback, still "down" since SO ~22

which RPS ?

it was r10960
works now again, after 61h ...

you put it in rescue without mail notification

[oles@ovh.net]

pendulum a écrit:
>
> After some fiddling around, I think the free and htop memory usage stats
> are so low because a lot of my programs use cached memory. I just
> wondered because it dropped from >100mb to that 35mb.

can you give the output of "free" of your RPS ?

[oles@ovh.net]

pendulum a écrit:
>
> I know that. But why not let the manager change both IP reverse dns
> entries or let the IP be the default that is in the manager? It's not a
> real problem but could cause confusion for people that don't know this.

hmm ... yes why ? ... I will check it.

[oles@ovh.net]

MDGeist a écrit:
>
> Feedback, still "down" since SO ~22

which RPS ?

[gunnarh]

IO/Wait when using Gallery2 on my RPS is much better this Week. Sometimes working on the console feels strange (seems to hang ~10 seconds when saving a config-file in vi but this behavior was yesterday - not today).

[pendulum]

After some fiddling around, I think the free and htop memory usage stats are so low because a lot of my programs use cached memory. I just wondered because it dropped from >100mb to that 35mb.

[F4RR3LL]

As Thomas, iv`e no problems with outgoing mails. Works fine for me.
No problems with reverse DNS. Little log for pendulum. Outgoing IP = incoming IP no Problem.

Mar 25 16:36:25 paulpanzer postfix/smtpd[4869]: NOQUEUE: reject: RCPT from **.de[87.98.183.***]: 450 4.7.1 <**.de[87.98.183.***]>:

My RPS works without any errors since yesterday afternon.
But working on the shell feels very slow sometimes.
I`ve got installed debian / lighttpd / mysql / php5 / postfix / dovecot / monit / backup2l an some own scripts.
Everything works.
Yesterday my RPS was ReadOnly from one second to the other.
I hope you`ve fixed this Problem in your big workout yesterday
So i think... if everyone here ist going on testing, finding bugs and you`re fixing it, its going to be a great cheap produkt.
regards Sven

//@Thomas kommen meine mails ned durch?

[pendulum]

I know that. But why not let the manager change both IP reverse dns entries or let the IP be the default that is in the manager? It's not a real problem but could cause confusion for people that don't know this.

[sledge0303]

I know but the IP that is used in the manager for the reverse dns is not the IP that is used for outgoing traffic. One has to change the default route to not run into problems with mailservers etc.

Mail is out.

Bind all your running services to that fail over IP, that's all. It works for me, it works for F4RR3LL...

[pendulum]

yes. it's normal. 91.121.196.254 is the router.

I know but the IP that is used in the manager for the reverse dns is not the IP that is used for outgoing traffic. One has to change the default route to not run into problems with mailservers etc.

Mail is out.

[sledge0303]

pendulum a écrit:
>
> Up to yesterday it was horrible. Crashes, timeouts, slow discs. To make
> a long story short: it wasn't useable for anything.
> But today it's a LOT better. Not a single crash up to now. Not even a
> dropped iscsi task!
> Speed is ok. Could be better but for the price it's enough.

We made lot of work on the iSCSI yesterday:
http://travaux.ovh.com/?project=15&s...all&perpage=50

> Problems that are still there:
> - my RPS seems to report wrong memory stats. (free, htop etc.)

which RPS ?

> - wrong IP routing. The RPS use their ISCSI IP for all outgoing traffic

!?

> -> reverse dns useless

do you mean: you can't change the reverse ?

Nope Oles,

as for me: Reverse DNS is working fine, MTA sends all mails as expected/configured. Peoples have to configure their MTA right now, that's all. Same with all other running services like webserver...
Just some words about the RPS: the intention with RPS as replacement for Kimsufi is nice. I have experienced a couple of time outs, working at console sometimes impossible, it takes some time till commands will be executed. top about 1 minute, working at console after login 2-3min...
Example these commands:

emerge --sync;emerge DNu world

Kimsufi needs 1,5 hours to compile all these packages and on RPS it takes about 3 hours.

[MDGeist]

Feedback, still "down" since SO ~22

So i now have 60 hours downtime...


Asked support already if i should reinstall (although i kind of dont want) or not...

[oles@ovh.net]

pendulum a écrit:
>
> The manager let's me change the reverse dns for the IP that is not used
> for outgoing traffic by default.
>
>
> Code:
> --------------------
> # ip route
> 91.121.196.0/24 dev eth0 proto kernel scope link src 91.121.196.xx
> 127.0.0.0/8 dev lo scope link
> default via 91.121.196.254 dev eth0
> --------------------
>
>
> The manager IP is 91.121.59.xxx
>
> Outgoing traffic will be routed through 91.121.196.xx

yes. it's normal. 91.121.196.254 is the router.

>
> Regarding the memory:
>
> free and htop say that 36mb are in use but I have mysql, php and other
> stuff running that I KNOW at least consume 100mb. I can write you which
> RPS it is by mail if you want.

oles@ovh.net

[pendulum]

The manager let's me change the reverse dns for the IP that is not used for outgoing traffic by default.

# ip route
91.121.196.0/24 dev eth0  proto kernel  scope link  src 91.121.196.xx 
127.0.0.0/8 dev lo  scope link 
default via 91.121.196.254 dev eth0

The manager IP is 91.121.59.xxx

Outgoing traffic will be routed through 91.121.196.xx

Regarding the memory:

free and htop say that 36mb are in use but I have mysql, php and other stuff running that I KNOW at least consume 100mb. I can write you which RPS it is by mail if you want.

[oles@ovh.net]

pendulum a écrit:
>
> Up to yesterday it was horrible. Crashes, timeouts, slow discs. To make
> a long story short: it wasn't useable for anything.
> But today it's a LOT better. Not a single crash up to now. Not even a
> dropped iscsi task!
> Speed is ok. Could be better but for the price it's enough.

We made lot of work on the iSCSI yesterday:
http://travaux.ovh.com/?project=15&s...all&perpage=50

> Problems that are still there:
> - my RPS seems to report wrong memory stats. (free, htop etc.)

which RPS ?

> - wrong IP routing. The RPS use their ISCSI IP for all outgoing traffic

!?

> -> reverse dns useless

do you mean: you can't change the reverse ?

[pendulum]

Up to yesterday it was horrible. Crashes, timeouts, slow discs. To make a long story short: it wasn't useable for anything.
But today it's a LOT better. Not a single crash up to now. Not even a dropped iscsi task!
Speed is ok. Could be better but for the price it's enough.

Problems that are still there:
- my RPS seems to report wrong memory stats. (free, htop etc.)
- wrong IP routing. The RPS use their ISCSI IP for all outgoing traffic -> reverse dns useless
- monitoring useless when ISCSI crashes as the box still pings but can't be reached through ssh

[oles@ovh.net]

Hi There,
What is you feedback about the RPS ? Have you any issue ? The problems ?
Does it work ? Doyou have any timeouts ? Is it slow ? Do you have any
crash ? The RPS is the worst solution that you have ever tested ?

Thansk for your feedbacks

We have an issue with 1 kind of using of the RPS that we think to
fix soon. Yesterday, we fixed lot of littel problems, but all isn't
finished. It has to be perfect. It's not yet. Thanks to help us.

Regards,
Octave