New SoYouStart 2018 Prices

jaypeesmith

I noticed that the order form still shows a $0 setup fee. > @Falzo said:

@AlwaysSkint said:
Kudos @TheLinuxBug & @Falzo !
Keep up the good work. :-)

thanks, but I have to admit I got stuck on trying to compile their kernel from their sources big time. luckily @linuxthebug knows people who are more experienced esp. on ARM ... let's see where this whole thing brings us. I am very curious myself ;-)

I would hope that, after the work you and @TheLinuxBug are doing to fix their issue, SoYouStart (OVH) would be grateful and offer you some sort of compensation (free hosting credit...wink wink).

Shot2

Why would they compensate, and what for? They did nothing wrong, that was and will be the official stanza. Remember OVH's moronic explanation: essentially, "the bug is ARM's fault, and anyway there is no bug".

jaypeesmith

@Shot2 said:
Why would they compensate, and what for? They did nothing wrong, that was and will be the official stanza. Remember OVH's moronic explanation: essentially, "the bug is ARM's fault, and anyway there is no bug".

Let me clarify. If someone is able to provide a solution and that solution can be leveraged to fix an issue for one's customers, it would not be unthinkable to show some gratitude for those who provided the fix.

I do agree that the SoYouStart's response was piss-poor. No one can tell me that they didn't know this was a issue before the roll-out.

TheLinuxBug

jaypeesmith said: Let me clarify. If someone is able to provide a solution and that solution can be leveraged to fix an issue for one's customers, it would not be unthinkable to show some gratitude for those who provided the fix.

To be honest, if we can make this work and come out with documentation that will help others to compile a working kernel as well, I will probably post a donation link for you guys to donate to the developer who spent the bulk of the time working on this and getting it to work! I think it would be fair to donate a few $ for his time if we find a working solution.

I hardly expect OVH will step up and take any responsibility, their position so far has been to act like they were being overly gracious by providing us the things they are mandated to by license (open source software). Had they not at least provided what they have so far, they would have likely been in violation of the license. So the way I see it, they have in their minds already fulfilled their responsibility to us, even if in a half-assed way. Not saying its right or wrong, just saying what the case likely is.

Cheers!

jaypeesmith

@TheLinuxBug said:

Well, at any rate, I appreciate the work. I'd have no problem donating as a "thank you" for the time and efforts put toward resolving this issue.

Falzo

TheLinuxBug said: I will probably post a donation link for you guys to donate to the developer who spent the bulk of the time working on this and getting it to work!

haha , your man is really pushing it. SYS support probably is already screaming at the amount of automatic monitoring mails they get for the 'stuck at reboot' moment :-D

Jona4s

Don't forget to try 4.4 kernel, as I mentioned it gives x5 performance than 4.9, on Arm machines.

KotCzarny

hello, after some hacking around i've come to following conclusions:

1/ all kernels compile/work just fine (tried 4.4.x/4.9.x/4.14.x/4.18.c),
2/ there is a regression in 4.9+ that results in awful upload speeds,
3/ quick hack helped to go from 4.6mbit to 1.6gbit with 4.9.119.

more work needs to be done. cheers!

mtsbatalha

@KotCzarny said:

3/ quick hack helped to go from 4.6mbit to 1.6gbit with 4.9.119.

more work needs to be done. cheers!

Can you please share the details?

KotCzarny

@mtsbatalha: not yet because it's quick and very dirty code hack

Jona4s

Just say the line of code you changed lol

Shot2

@KotCzarny said:
2/ there is a regression in 4.9+ that results in awful upload speeds,

Did you locate, and report the regression?

KotCzarny

@Shot2: nope, at the moment i'm trying to have something working and stable.

anyway it's a pity that 4.14+ had mvpp2 driver redesign resulting in nonworking packet flow (probably something simple i have yet to fix). but there is a chance 4.9.119 would be working and be good enough [tm] for most people because it's a longterm kernel series

sin

Thanks again guys for working on the kernel :-)

TheLinuxBug

@KotCzarny , @Falzo and I have been at work testing things:


Some more teasers:





Iperf Test (In KBytes/sec):



luks encrypted volume working:

We are still doing beta testing for the kernel to make sure there are no issues with the 'fixed ' network driver. Until we know our changes are stable, we will be doing internal testing for now. Once we have confirmed things are stable and working and are ready for public testing, we will start a new thread to post about it.

That said,if you are serious about testing, willing to risk your data and can be on irc to provide feedback while you test, come to Freenode on #sysarm and we will pick a few people to help.

Cheers!

Shot2

You should try luks in aes-cbc[-essiv:sha256] mode, too. It's quite fast (ARM-accelerated) with the 4.5.2 kernel... but crashes when in use.

TheLinuxBug

Shot2 said: You should try luks in aes-cbc[-essiv:sha256] mode, too. It's quite fast (ARM-accelerated) with the 4.5.2 kernel... but crashes when in use.

Once step at a time but maybe get with Falzo and when he is done/with his help give it a go. If this one wasn't production and the other being used as our dev it would be a different story ;p

Anyone who wants to test can come to #sysarm on Freenode and we will pick a few serious people to help.

Cheers!

AlwaysSkint

Trying HexChat for the first time...

AlwaysSkint

This is looking promising..

Linux sys 4.5.2-armada375 #1 SMP Tue Oct 25 11:52:56 CEST 2016 armv7l GNU/Linux
root@sys:~# iperf -c iperf.ovh.net -i 1 -t 15 -f KBytes

[ 3] 0.0- 1.0 sec 157056 KBytes 157056 KBytes/sec

Linux sys 4.9.119-kc375 #6 SMP Sun Aug 12 09:58:26 CEST 2018 armv7l GNU/Linux
root@sys:~# iperf -c iperf.ovh.net -i 1 -t 15 -f KBytes

[ 3] 0.0- 1.0 sec 204416 KBytes 204416 KBytes/sec

Got a problem with iptables not liking the modules, even after soft-linking to the existing 4.9.58 ones.

modprobe: ERROR: ../libkmod/libkmod.c:586 kmod_search_moddep() could not open moddep file '/lib/modules/4.9.119-kc375/modules.dep.bin'

4.9.119-kc375 -> 4.9.58-armada375

modprobe: ERROR: could not insert 'ip_tables': Exec format error

iptables v1.6.0: can't initialize iptables table `filter': Table does not exist (do you need to insmod?)
Perhaps iptables or your kernel needs to be upgraded.

TheLinuxBug

AlwaysSkint said: Got a problem with iptables not liking the modules, even after soft-linking to the existing 4.9.58 ones.

You obviously didn't pay attentions to the files provided (facepalm). The Modules are provided in a tar.gz... You should instead extract them in /lib/modules and then run depmod -a......

BTW if you would have communicated that on IRC we would have told you

Cheers!

AlwaysSkint

D'oh - I was too quick to test and didn't even look at the .gz

Epic fail.
((Goes off to the dunce's corner))

peu_importe

@KotCzarny @falzo @TheLinuxBug
hello, i have 2 sys arm lying arround that i am willing to lend to hasten solving this kernel problem.
if that help, you can send me a dm for credentials

AlwaysSkint

Looking pretty stable to me, on the beta kernel..
An rsync via webdav of cPanel backups, from a USA VPS:

sent 10,968,171,294 bytes received 1,586 bytes 2,849,246.10 bytes/sec

total size is 16,405,206,175 speedup is 1.50

real 64m21.617s

user 1m38.213s
sys 0m35.728s

It saturated the server with a Load between 1.75 and 2 for the duration, which is fine for the secondary (encrypted with Nextcloud) backup that this is being used for.

KotCzarny

@AlwaysSkint: grab new modules pack or compile cryptodev and check if you can make sshd use hwcrypto?

KotCzarny

btw. for the lazy ones i've put openssl compiled with cryptodev support in hopes anyone can convince sshd to use hw crypto

AlwaysSkint

@KotCzarny updated, thanks. Impressed with latest iteration so far and will continue to monitor/test.

Dwayne

@Neoon said:

@Harambe said:

@Dwayne said:

Well, my German SYS login details get rejected on the /ca/en page and I am unable to change the country when trying to create a new account on the /ca/en page:

Worldwide $ site: https://www.soyoustart.com/us/server-storage/

or Worldwide € site: https://www.soyoustart.com/ie/server-storage/

Will need to create a new account, and potentially re-verify

Oh yea, fuck, my fault then ca is not worldwide, I changed it to us now.

My fault, fixed.

Managed to snatch one a few days ago. Thank you @Neoon

AlwaysSkint

Anyone got ipset working? If so, which kernel and any gotchas?

root@sys:~# ipset create blacklist hash:ip counters family inet maxelem 2000000

ipset v6.30: Kernel error received: Invalid argument

Used in csf:

IPSET: [ipset v6.30: Kernel error received: Invalid argument]

iptables v1.6.0: Kernel module xt_set is not loaded in.

citrix

i see SoyouStart now has servers in .DE as well

_New Essential server regions
Discover our new Dedicated Servers
From £35.99 ex. VAT/month (or £43.19 incl. VAT) _

... also a note says soon UK

https://www.soyoustart.com/en/essential-servers/

TheLinuxBug

@AlwaysSkint said:
Anyone got ipset working? If so, which kernel and any gotchas?

root@sys:~# ipset create blacklist hash:ip counters family inet maxelem 2000000

ipset v6.30: Kernel error received: Invalid argument

Used in csf:

IPSET: [ipset v6.30: Kernel error received: Invalid argument]

iptables v1.6.0: Kernel module xt_set is not loaded in.

Maybe you should ask for the modules to be compiled? We only built things in that were deemed needed at the time, obviously once we release all the information you can build the whole kernel and whatever modules you want from scratch. Also, it would be much appreciated if you report the things on IRC and wait around so we can help instead of posting here -- most people won't even know what your talking about and you will get much quicker feedback on IRC as well.

Cheers!