VMs Network speed slowing down

[amil]

This is a limitation of OVH's connection. I got the same results on Servers under 100 Eur/month.
They say that they don't throttle - but they do. Really Painful to backup my data every day.

In the other direction 100Eur+ Server -> cheap Server is no limit. Maybe anybody can confirm that !?

I think that this is not correct, since i had servers with soyoustart and the subnet worked as expected without limitations, so i don´t think that they do limitations on subnets "by default" on their main brand with a server with 3x the cost.

If i got a dedicated server (with any company) i understand that subnets are routed, so i should receive similar results in ping/tracert/ and up/down speeds. So if hardware node is overloaded of course the other IPs on server will affect, but is not absolutely the case

@PhilV any good news regarding this? Did you have any OVH update?

 

[amil]

I have rebooted server in rescue mode and configured in eth0:0 one IP from subnet to make test (without vmac assigned, with vmac not works here).

And it seems that a simple wget --bind-address= it gives good results. The same stable speed and continuous over the time.

Another track: in a VM with 5 extra IP (from subnet), with a "shared" vmac the speed is the worst of all VM, like they are limiting via vmac and if you share vmac, more limits.

Anybody can confirm this please?

 

[PhilV]

@PhilV any good news regarding this? Did you have any OVH update?

Last night OVH IT Expert took hold of the server for testing... "pro" rescue mode they said...
Waiting for debriefing

 

[d1ch0t]

Thought I'd report back on this. After running some "tests" OVH stated there was no issue their end. I then pointed out that was unlikely given this thread and that one of their support staff had already reported there was a problem with their network. At this point they went dark.

The next day I ordered a new box in RBX which has none of these issues and asked for a cancellation and refund on the GRA box. They refunded me.

 

[amil]

@PhilV any good news from your case?

In my case the tecnicians it supposed they are checking the issue (and have the permission to access server via ssh to make test) but i have no news. Their previous checks was ok for ovh.

@d1ch0t so, in your case they didn´t found the issue and you had to change datacenter. It seems that they have this issue in their datacenter in Gravelines.

 

[jacobwilliam]

A VM has access to practically the same Internet speed as the host PC; no more of a slowdown than going through a typical router.

If you have other programs running in the VM that are using up ram and causing the pagefile to thrash, it will effect the speed that the download can get written to disk, thus slowing the download. It's proabably a long shot, but worth investigating... unless I misunderstood and you mean web pages are slow downloading. If this is the case, the amount of ram would probably be much less of an issue if any.

 

[PhilV]

@PhilV any good news from your case?

No good news at this time...
OVH Support ask me for the second time to access the server... They want to make Iperf Test from the IPFOs in rescue mode.
I am not optimistic...

 

[PhilV]

A VM has access to practically the same Internet speed as the host PC; no more of a slowdown than going through a typical router.

If you have other programs running in the VM that are using up ram and causing the pagefile to thrash, it will effect the speed that the download can get written to disk, thus slowing the download. It's proabably a long shot, but worth investigating... unless I misunderstood and you mean web pages are slow downloading. If this is the case, the amount of ram would probably be much less of an issue if any.

As I said, there is no other program in the the VM : just create a CT (or VM KVM), add template Debian 8.4, that's all...

When starting VM, network is full speed equal to host... after few minutes, download fall down.
It's not penalizing for a little apt-get upgrade of 12 MB, but unusable for recieving a 25 GB backup from an other server, or syncing a large database...

 

[amil]

@PhilV thank you for your reply, hoping the best for you.

jacobwilliam: "A VM has access to practically the same Internet speed as the host PC; no more of a slowdown than going through a typical router." -> This should be the case of course, but this is not our case, and it seems there are many cases, all in Gravelines datacenter and it appears they have no idea at this time why is happening.

Since this Friday, no news in my case.

@PhilV i can confirm that if they make test in rescue will work well for them. The issue (i am 99% sure) is when we are forced to use vmac on virtual machines to get working with bridge. Here comes the bad.

"When starting VM, network is full speed equal to host... after few minutes, download fall down." -> Exactly the same case of me.

This was my iperf test on VM:

iperf -c iperf.ovh.net -i 2 -t 10 -r
------------------------------------------------------------
Server listening on TCP port 5001
TCP window size: 85.3 KByte (default)
------------------------------------------------------------
------------------------------------------------------------
Client connecting to iperf.ovh.net, TCP port 5001
TCP window size: 196 KByte (default)
------------------------------------------------------------
[ 5] local 13*.**.**.** port 34635 connected with 188.165.12.136 port 5001
[ ID] Interval Transfer Bandwidth
[ 5] 0.0- 2.0 sec 219 MBytes 919 Mbits/sec
[ 5] 2.0- 4.0 sec 225 MBytes 942 Mbits/sec
[ 5] 4.0- 6.0 sec 225 MBytes 943 Mbits/sec
[ 5] 6.0- 8.0 sec 224 MBytes 939 Mbits/sec
[ 5] 8.0-10.0 sec 225 MBytes 944 Mbits/sec
[ 5] 0.0-10.0 sec 1.09 GBytes 937 Mbits/sec
[ 4] local 13*.**.**.** port 5001 connected with 188.165.12.136 port 43838
[ 4] 0.0- 2.0 sec 2.09 MBytes 8.75 Mbits/sec
[ 4] 2.0- 4.0 sec 1.57 MBytes 6.60 Mbits/sec
[ 4] 4.0- 6.0 sec 1.51 MBytes 6.34 Mbits/sec
[ 4] 6.0- 8.0 sec 2.19 MBytes 9.19 Mbits/sec
[ 4] 8.0-10.0 sec 1.52 MBytes 6.39 Mbits/sec
[ 4] 0.0-10.5 sec 9.12 MBytes 7.31 Mbits/sec


Not good.

@PhilV If you want, suggest then that they make test -in rescue mode- but setting/unsetting the vmac on the same IP. Of course i had notified of this, all my test and no response, and i was unable to get working in rescue a subnet IP with the vmac assigned.

While in rescue mode, it appears okay. Of course the additional IP i used for test was without vmac, and this is the reason the IP was stable and full speed.

So, all points to (supposed/i want to think) unintentionally the bw is being limited via vmac. In the case in one VM has multiple IP -sharing vmac- the total "limited" bw is divided for each IP, so bw is very very worst in this cases.

As previously comment @d1ch0t in his case, they refund and he goes to RBX where it is supposed he don´t have the problem.

So no very good news, since they appears that they not find -or do not share, or do not assume- the root issue. But yes, is happening in GRA.

I can´t put this server in production and we can´t assume this.

I´ll keep updated post. Good luck! Greetings!

 

[PhilV]

@amil , I agree with you... at 99% there's a pb with vMACs.

And not Only in GRA Datacenter, the 1st server I got was in SBG-1 Datacenter... same issue.

For now, OVH support is testing in rescue mode, but I do not dream

I suggested them to provide me a usable server with Proxmox (I really enjoy it) regardless of the Datacenter as it is in France...
Waiting again... that's 2 months now.

 

[LordY971]

Did you tried to change the network interface type of your VM? E1000, vmxnet3 or virtIO Eternet card ... Normally E1000 is sufficient !

 

[PhilV]

Did you tried to change the network interface type of your VM? E1000, vmxnet3 or virtIO Eternet card ... Normally E1000 is sufficient !

thanks for answer

Yes I tried differents types of network Model... no luck.
Additionnally, I prefer use LXC Containers, so it's not concerned by network type like KVM are I think.

 

[amil]

Did you tried to change the network interface type of your VM? E1000, vmxnet3 or virtIO Eternet card ... Normally E1000 is sufficient !

Yes, here also try with different virtual interfaces (virtio, e1000), and also with LXC containers, and the same issue. It is not a proxmox or configuration issue.
@PhilV any good news?
Thanks for confirm that this issue is not only on GRA datacenter.

 

[amil]

Not good news for me. They not found any issue in all test that they supposed to do.
I asked for their tecnicians setup a rescue mode with a subnet IP attached to their vmac, still no response.

I am only with 5 days to renew and i am not going to renew with this problem, the server is totally unusable.

@PhilV any good news in your end?

 

[extor]

Hello!

Thanks for posts!

I was looking for the problem of the whole day, until I found this forum.

We have same issue and do several tests.
VMs network speed slowing down only with external network. Speed is ok between VMs on 2 KVM hosts.

 

[PhilV]

@PhilV any good news in your end?

No good news unfortunately...

Last message from OVH Support :

demand is still being checked with administrators Network.
So either the concern is related to a saturation of our network or related to the software configuration of your virtualization system.

I wait for 37 days... I losing patience.

 

[PhilV]

Hello!

Thanks for posts!

I was looking for the problem of the whole day, until I found this forum.

We have same issue and do several tests.
VMs network speed slowing down only with external network. Speed is ok between VMs on 2 KVM hosts.

Hello extor,
Do you have open a support ticket ?
More we will, more they will take the problem seriously !

 

[extor]

@PhilV, yes of course.
I think exactly the same, so I post a message.

 

[amil]

@PhilV i´m sorry for the bad news. In my case bad news also, my server is due to renew and of course i don´t go for renew. I am going to give another oportunity, i am going to order a new server and i will try again.

It is very sad that a "big" company is omitting our complaints and simply "its our fault".

@PhilV will you get another server or you will wait for any good resolution? Did you request a refund?

 

[PhilV]

@PhilV
@PhilV will you get another server or you will wait for any good resolution? Did you request a refund?

Hi,
This is actually the second server ! The only thing the two servers has in common is... the IPFOs/vMACs...
That's why I pretty sure this is the source of problem...

Anyway,
If anybody here have an OVH server with Proxmox 4 fully fonctionnal (network part regarding this thread) and can share his network conf for host/CT, maybe it could help !