hi
i ordered a VPS yesterday
maybe 1 - 2 PM EST
hm wait
never mind
phlux: color for wifi strength?!  that's gangster right there...
brycec: oh cool, thanks for the info (re. mSATA and the OCZ Agility)
up_the_irons: re kvm-clock as tsc; are vm's encouraged to run their own ntpd or can we trust the host?
jbergstroem: i think you can trust the host but doesn't hurt to run your own.  almost all hosts run ntpd, but there may have been some in the beginning that didn't
up_the_irons: noted, thanks.
np
up_the_irons: I ended up disabling the network monitoring :(
I run ntpd on every host I have KVM or otherwise, matter of habit
staticsafe: but do you run two? :)
:P
up_the_irons: thanks for the contect... Had no idea what you were talking about :p
brycec: hah
given it was a few days ago, i figure some context would be nice:)
Wonder if there are server boards with mSATA... You could put the host on a cheap/small 20GB mSATA and leave all interfaces and drive slots for storage.
Exactly
Welcoem back
can't hotswap :/
what happens when it dies
Hi. I'm planning to order a VPS. Some questions: Does your "Out-of-band management" service has any web-based interface rather than the serial console one?
And if so, any demo page that I could browse?
deXar: no, the serial console one is over SSH, but there's also a VNC one (so any VNC viewer would be what you'd be looking at :)
i had played around with noVNC, to pop up a web-based out-of-bander.  It worked very well, but requires some setup on each and every VM, so unfortunately i need to make a project out of it
mercutio: hot-swapping the system drive? Doesn't make sense, unless the system drive were RAID. (and technically you could probaly hot-swap, provided you opened the chassis)
I'm thinking about coloring load averages in conky..
The "VNC viewer" is also text-based, right?
mercutio: basically, "what haoppens when [the system drive] dies"? Same thing as now... things keep running until it all goes to hell, in the meantime you migrate guests to other hosts :p
up_the_irons: Also, then how it's possible to mount an ISO image remotely in case of serial console? A full URL of the iso would be needed, or?
deXar: no, the vnc viewer is a GUI program.  to choose the media to boot (HD, CD-ROM, etc...), you need to use VNC; unfortunately the serial console does not redirect all the way up to the BIOS level, instead it redirects starting with your OS loading (if you configured your OS to do that)
up_the_irons: I see. Would the default base installation of OpenBSD support VNC Viewer GUI? - Planning to get an OpenBSD VPS and I just wish to mount my own OpenBSD ISO whenever I'd need, install it with custom partitions, etc.
deXar: all our VMs support the VNC viewer, it is a host machine thing, not an OpenBSD thing.  You still can't mount your own ISO, we need to pop it into your VPS virtual CD-ROM on our end (but we don't mind doing this whenever needed)
up_the_irons: I understand now. Nice to know that. Many thanks for your helpful answers. See you again. Regards.
You don't need the ISO to upgrade OpenBSD though.
You only need it if you don't have a bootloader.
Even then, any old OpenBSD ISO will do.
brycec: well i'd think you would have backups of root drive, and not much on it so you could reimage
right and right
but normally it's a pita to open up a computer compared to swap a drive at front
Hell, I'd even consider centralised/netboot for the host OS... but that introduces an extra level of dependency/failure
deXar: no problem!
brycec: i dunno about extra dependency being a bad thing in that case. I mean, if it's your nas / san hosting it I'm sure you'd be hosed anyway if it couldn't connect
brycec: if anything removing internal storage media is probably a good thing
brycec: i just hate the latency of running a machine over nfs / iscsi
brycec: in my experience it just isn't fun, but that's mostly noticed in sqlite databases for package managers, where they are bound by disk transaction latency for many small writes
wakIII: Depends on the setup, of course. In ARP's case, as I understand it, each machine is mostly-independent, has its own system drive and RAID array, doesn't rely on a central SAN. Probably a smart apporach too - no single host machine can affect the other host machiens, save for network congestion.
Not to mention keeping the VM's near-side keeps IO latency minimal.
oh yea
Of course up_the_irons just needs to replace all his disks with SSD's now...
i definitely would want optimized latency in this case
<.<
lmao
gonna need to do more than that to compete with vendors like linode
I dunno, I think ARP does plenty that's competitive with linode
(and others)
well, they were more competitive on the linux side until linode massively upgraded their offerings
But it depends on what you value, I suppose. Linode is like Walmart, and ARP is the local chain that offers value instead of sheer price.
yea maybe
Besides, last I knew (been awhile) Linode was Xen, so not very BSD-friendly
well that's the problem
I definitely prefer full VM to hypervisor-based. KVM ftw.
that's one of the reasons i came here in the first place
ehh, they could do xen hvm similar to kvm
but their setup revolves are pv
s/are/around/g
i was surprised to see just how performant xen hvm is these days, it's pretty much even with kvm
My only objection to Xen is that I'm still left at the host's whim, notably bootloader, and even the boot process, as well as the requirement to run a Xen-patched domU.
In a resold environment, that's just not useful to me
If i were running my own VM in-house, I would consider it...
But I prefer how clean and simple it is to say "here's your VM, it acts like it's bare-metal hardware"
but xen hvm does just that
it's just that most hosts go one way or the other. Some do hvm some do all pv
i dunno, i just use libvirt anymore so pretty much anything goes for full virtualization as long as it performs reasonably well