New on LowEndTalk? Please Register and read our Community Rules.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
Comments
Party crash the node with:
Loads of 250, here we go!
Haha, this is nice:
Lets do more...
load average: 314.03, 300.53, 284.51
Not as high as some, but quite good.
@DotVPS WTF. That's crazy, what are you running. I was running like 5 ffmpeg instances.
@Infinity
You didn't see my 10000? Muahahaha.
Btw now I can't access u_u Maybe I got locked :P or the whole node is locked XD
Has anyone ran linpack yet?
There was someone that forkbombed linpack back in the day :P
Francisco
@Francisco What would cause someone to trigger the fraud detection when signing up?
They detected you have no intentions to abuse your VPS?
Maybe however I doubt it. More likely that the IP you are registering with doesn't match the country you selected, that happend to me on Frantech before I think.
I wasn't using a proxy or anything, but geographically I think the IP my ISP assigns is halfway across the state.
@u4ia - Just PM me the email you signed up with, and I will take a look.
@Aldryic
sent, thanks!
@Francisco
I know its not really what this test is about but mine has IPv6 addresses but no actual connectivity in or out...
Was hoping to test if this kernel better supports ip6tables...
@hunterminogue - ipv6 is 'standard' for us. This is just load/stress testing. Production nodes are all IPv6 ready.
Our .18's support it for the most part. As far as i've seen ip6tables works on these .32's too
Francisco
Fran is there any chance of me getting to have a little private time with a pony? 3685701435 :P
@Francisco
I found the openvz .18 kernel doesnt seem to support stateful matching eg
$ ip6tables -vL
Chain INPUT (policy ACCEPT 0 packets, 0 bytes)
pkts bytes target prot opt in out source destination
0 0 ACCEPT all lo any anywhere anywhere
0 0 ACCEPT all any any anywhere anywhere state RELATED,ESTABLISHED
0 0 ACCEPT tcp any any anywhere anywhere state NEW tcp dpt:ssh
2 160 REJECT all any any anywhere anywhere reject-with icmp6-port-unreachable
Showing my ipv6 ssh attempts being silently rejected. I found only stateless matching worked.
according to http://www.sixxs.net/wiki/IPv6_Firewalling stateful inspection requires kernel >=2.6.20
On the test .32 box it wont even load eg
$ /sbin/ip6tables -A INPUT -m state --state RELATED,ESTABLISHED -j ACCEPT
ip6tables: Unknown error 18446744073709551615
Not sure if this is kernel related or a config problem with the container.
@Taylor - I'll see if Anthony has some spots
@hunterminogue - I'll put it to work on our private test node to see if that's the case.
To everyone else, it's looking like vswap is working now, or at least according to 'stress'. memtester still can't allocate the whole thing but when using stress with the following:
I get the following
Francisco
I'm pretty sure i've been able to track down the ip6tables stuff.
Please try it out but please be aware that:
Please let me know how it goes @hunterminogue
Francisco
Nope.
As I had documented, vswap is not perfoming as it should. To quote my bug:
http://bugzilla.openvz.org/show_bug.cgi?id=2102
We simply can't have a node going into real swap just because a user is using vswap. There's no reason the rest of a node should be dragged down from a single user. This sounds like it's a real design flaw in how vswap is being handled. We'll see if this is addressed in some newer kernels but so far it's appearing to be that way.
Francisco
which isn't good at all.
What is HN swap?
And I don't understand your comment lol, but yes, when I had my Uptimevps, it allocated swap despite having lots of free RAM (since start, as you said).
So, if I can understand something, the containers are pushing memory to the swap node without reason?
HN = hardware node, so the node the VPS is on
It seems that vswap is actually allocating real swap. Real SWAP will tank IO, especially if you get a decent amount of it in use.
At this point vswap is broken again after I told swap to turn off.
I'm going to just allocate things different and be done with it
Francisco
So, if no hardware swap + no vswap = happy node?
I think swap is useless
That's what we're aiming at.
I think vswap is a cool idea but it has way too many problems.
To date we've had no new issues with the nodes since we dealt with vswap & the forkbombs. If it wasn't for me taking down the nodes for some other work yesterday both nodes would be well into their 2nd week.
Francisco
To add to the vswap stuff.
When I tuned swappiness back to 60 I was able to allocate it without issue.
Francisco
LOL, weird things.
Don't swap... imho.
As I said, the new memory allocation is better and you can run more with less memory, so, letting people use swap just will slow other people. I think the guaranteed memory is more than enough.
As an example, I can run the same thinkgs that I run in your $15 plan and around 160MB in less than 100MB with .32
To stop the crying we'll likely offer 2x the RAM with half of it being considered burstable.
I'm glad to hear that It sounds quite promising.
Francisco