How To: Determining how many 'VPS neighbors' you have or if you are on an oversold OpenVZ node

[Awmusic12635]

One of our dev nodes:


[root@dev3 ~]# cat /proc/cgroups
#subsys_name    hierarchy       num_cgroups     enabled
cpuset  3       6323    1
cpu     3       6323    1
cpuacct 3       6323    1
devices 4       6322    1
freezer 4       6322    1
net_cls 0       1       1
blkio   1       6323    1
perf_event      0       1       1
net_prio        0       1       1
memory  2       6322    1

I think you win

 

[SentinelTower]

Wow, this is a nice piece of information. Is there a way for providers to hide this file? However this only gives us the number of containers, how can we know if the node is oversold ?

 

[Amitz]

Damn. I just thought "Cool, let me check that on my VMs!" until I realised that I only have XEN VMs and dedicated servers in the meantime... Not a single OVZ left. What a pity.

 

[tmzVPS-Daniel]

FINALLY! A way to prove clients that you do not over-sell. 

- Daniel

 

[Geek]

You're welcome.

 

[Kalam]

Hmm, apparently every single OpenVZ I have from multiple providers is running an older kernel.

 

[dabtech]

XVM Labs

root@la:~# cat /proc/cgroups
#subsys_name	hierarchy	num_cgroups	enabled
cpuset	2	583	1
cpu	2	583	1
cpuacct	2	583	1
devices	3	582	1
freezer	3	582	1
net_cls	0	1	1
blkio	1	583	1
perf_event	0	1	1
net_prio	0	1	1

 

[Fusl]

Better don't host anything on my test host node at home:


[root@chernobyl:~] vzctl exec $(vzlist -1 | head -1) "cat /proc/cgroups"
Executing command: cat /proc/cgroups
#subsys_name hierarchy num_cgroups enabled
cpuset 3 19275 1
cpu 3 19275 1
cpuacct 3 19275 1
devices 4 19274 1
freezer 4 19274 1
net_cls 0 1 1
blkio 1 19275 1
perf_event 0 1 1
net_prio 0 1 1
memory 2 19274 1

And I wonder why it's so really f***ing hot there...

 

[William]

This test should work with newer versions of LXC and Linux- VServer (used by Edis) too shouldn't it?

No, it won't work on Vserver and never will. OpenVZ only. No LXC either.

 

[HalfEatenPie]

Wow, this is a nice piece of information. Is there a way for providers to hide this file? However this only gives us the number of containers, how can we know if the node is oversold ?

I'm pretty sure general logical reasoning can be applied here.

Assuming how beefy the host node is (if they tell you that is...), what kind of configuration they have, you can roughly assume if it's oversold or not. 

As a general rule of thumb, an E3 node should not have 200 VPSes on it.  Or in matthew's case, several hundreds on a single hard drive setup >.>

Better don't host anything on my test host node at home:


[root@chernobyl:~] vzctl exec $(vzlist -1 | head -1) "cat /proc/cgroups"
Executing command: cat /proc/cgroups
#subsys_name hierarchy num_cgroups enabled
cpuset 3 19275 1
cpu 3 19275 1
cpuacct 3 19275 1
devices 4 19274 1
freezer 4 19274 1
net_cls 0 1 1
blkio 1 19275 1
perf_event 0 1 1
net_prio 0 1 1
memory 2 19274 1

And I wonder why it's so really f***ing hot there...

That...  What do you even run locally that requires that many OpenVZ VPSes?!!!!  I can't even find a reason to run more than 10 VPSes locally!

 

[Fusl]

That...  What do you even run locally that requires that many OpenVZ VPSes?!!!!  I can't even find a reason to run more than 10 VPSes locally!

OpenVZ w/ ploop on an NFS4 mount - Functional and performance testing for the love of OpenVZ

 

[dcdan]

OpenVZ w/ ploop on an NFS4 mount - Functional and performance testing for the love of OpenVZ

If you don't mind me asking, what OS template were you using?

Also, how many processes do you see on the host node? (ps aux | wc -l)

Thanks

 

[Mid]

OpenVZ w/ ploop on an NFS4 mount - Functional and performance testing for the love of OpenVZ

testing 19273 times !

Sorry to say, but please check with a doc whether you have 'OCD'. Seriously.

 

[Mid]

If you don't mind me asking, what OS template were you using?

Also, how many processes do you see on the host node? (ps aux | wc -l)

Thanks

I am not a admin or hosting guy, but just a casual user 

The os should be debian/ubuntu/cent, asking for reliability?

processes must be 19273+x  where x < 100 

 

[dcdan]

I am not a admin or hosting guy, but just a casual user 

The os should be debian/ubuntu/cent, asking for reliability?

processes must be 19273+x  where x < 100 

Well there are two reasons why I am curious:

1) Process count. After about 50000 processes on the host node it will start locking up. If each VPS runs 3 processes (absolute minimum, basically just init and two kernel processes) that's already almost 60000.

2) Even a minimal centos install times 19273 equals 10 TB of data

 

[HalfEatenPie]

I'm 100% sure @Fusl knows what he's doing  I mean it was even said that it was for performance testing, so obviously you're going to see some people trying to push the boundaries outside of normal parameters.  

 

[dcdan]

I'm 100% sure @Fusl knows what he's doing  I mean it was even said that it was for performance testing, so obviously you're going to see some people trying to push the boundaries outside of normal parameters.  

By no means I implied otherwise. Just genuinely curious.

 

[MannDude]

Ah, looks like they're going to treat this as a bug instead of a feature: https://bugzilla.openvz.org/show_bug.cgi?id=3234

 

[drmike]

Ah, looks like they're going to treat this as a bug instead of a feature: https://bugzilla.openvz.org/show_bug.cgi?id=3234

Shame, what sorry sports.  It was an honesty feature that made at least one piece of data transparent on providers servers.

Can't have that sort of thing happening  Must mask the real values... 

What's wrong, a bunch of providers ashamed of their load numbers getting out in public?

 

[MartinD]

Could always overcome it with a level of slabbing.