https://www.arpnetworks.com/order?product=vps&vps_special=1
I added a "Tip"
I wonder if this will increase orders for those who don't know you don't have to stick with what we give you
In my case you were cool/smart/awesome/wise enough to include the right OS, so I didn't have to load it myself. :)
:)
Hello, up_the_irons!
hi dxtr
What's up?
up_the_irons: eeek, and I can't actually connect to the server
oh wait... could be an IPv6 issues on my end
actually....
Connecting to 2607:f2f8:0:102::3:80... failed: Connection refused.
up_the_irons: FYI ^^
doing a dist-upgrade on that server :)
up_the_irons: ahhhh :)
up_the_irons: You know what would be cool?
up_the_irons: oh yeah, did you see that KVM bit I sent before about CDRom changes?
G: yeah, but my interaction with kvm needs to go through libvirt, not to qemu directly
or else changes are lost after a reboot
up_the_irons: but newer libvirt/KVMs, allow hotswapping hardware (and It hink CDs too)
G: if i can get a newer libvirt on jaunty, i'd have this problem solved already; b/c newer libvirt has more features (like you said).  last time i tried this, i got caught up with udev needing to be newer (= dependency hell), but i resolve to try again
up_the_irons: ahhhh yeah, udev dependency nightmares
major pity then
yeah
time for an upgrade then ;)
actually, roughly when was Jaunty released?
2009 right?
don't remember
up_the_irons: what is the out of the box libvirt version, (I don't have a handy copy)
0.6.4-1arpnetworks
i use a custom one
a bit newer
for jaunty at least
hmmmm w/ udev 095 (plus I guess some patches), Red Hat has managed to get libvirt 0.8.1 working
(this is RHEL5, which was released early 08
yeah
up_the_irons: checking out what Red Hat did to get it to work might be helpful, http://ftp.redhat.com/redhat/linux/enterprise/5Server/en/os/SRPMS/ is where they hide all the sources
i'll check it out if needed
i'm back!
i'm one of the customers trying to get plan 9 installed and working at arpnetworks. one of the other customers has told me they arranged to have their boot floppy image swapped out by a cron job. (we've been testing different boot floppies with different configuration options). by chance, is the support staff member who arranged this present in the channel?
_pe: best to email support@ ;-)
_pe: is kvm different than running qemu current? I would presume it to be close enough that instead of paying for a vps to hopefully make it work you'd create a vm image and/or an install procedure that you know works ..
toddf: there are some peculiarities to arpnetworks' setup, yes. i don't have amd64 hardware to play with, locally, but other people are running plan 9 in kvm/qemu on ubuntu and i run plan 9 in qemu on openbsd.
ah. can plan9 network boot? might be easier to get two vps's and run qemu on one to simulate a plan9 boot server etc unless it can boot off services provided by a std unix system
the problem is the kernel is not getting very far before it halts. we think we know where the problem is, but testing requires some trial and error.
the user who has the cron job setup actually got it to boot and did an install but his boot method seems to have disabled some essential drivers.
we hope to develop a generic install method so garry can add plan 9 as a supported system.
sweet
whoa
plan9 on arp? neat.
greetings
any freebsd zfs folks out there?
bharatak: Yes?
pilgrimd, i'm confused about the tuning, lots of contradicting posts. i've got 4G but my vfs.zfs.prefetch is still disabled and getting some poor speed
going to enable it and reboot
but - dont really want to play with arc size, etc
Are you running 32- or 64-bit?
Tuning requires you establish a baseline kmem and arc usage level for your system, but there are some initial tuning guidelines I can give you.
There's a lot of conflicting info because a lot of people got really happy about ZFS on FreeBSD and were performance tuning it and working around the problems of making ZFS work on a 32-bit OS.
64bit
On 64-bit, you pretty much don't need to to any tuning except vm.kmem_size and vfs.zfs.arc_max.
pilgrimd, if you can point me to some docs, that would be great
i'm moving away from openindianna/solaris where it all worked
good move imnsho
pilgrimd, that sounds reasonable, limit arc so it dont take all, in solaris arc would give back memory when the system needed it
I don't have a doc that says this, just what I've gleaned in the last few years.
a_j__c: yeah, not only do i get security updates, I get software ;)
:)
pilgrimd: should i enable the prefetch?
what does that actually do
bharatak: How much memory do you have in your system?
4G
I'd leave it disabled.
lightly used, just a personal web/email and a file server for 2 clients or so
You really should get more memory. :)
Run this command in bash: echo $(($((`kldstat | awk 'BEGIN {print "16i 0";} NR>1 {print toupper($4) "+"} END {print "p"}' | dc` + `vmstat -m | sed -Ee '1s/.*/0/;s/.* ([0-9]+)K.*/\1+/;$s/$/1024*p/' | dc`))/1048576))
i know, system is a bit old, uses ddr2 memory and thats gotten $$
The number printed is the amount of kmem (in MiB) your system is using.
It might also work in regular bourne shell as well.
pilgrimd: not much, 1884 - but i'm accessing remote not using x, etc
On a 4 GB system doing not a whole lot, vm.kmem_size="3072M" and vfs.zfs.arc_max="1536M" is reasonable.  It depends on if you have UFS filesystems as well and how big your userland needs to be.
Ok, then vm.kmem_size="3072M" and vfs.zfs.arc_max="1024M" should work.
You'll want to do more profiling over time.  Keep an eye on kstat.zfs.misc.arcstats.size and see where it settles.
the default settings are kmem ~4100M, arc ~ 3000
Right, and with 4GB system memory, you'll run into swapping.
The whole point is to limit kernel memory to prevent swapping, which is murderous on a ZFS system because of the ARC and ZIL.
so kmem, is the total memory the kernel will use? what about userland apps
Solaris has this really great livelock deal where you can run it out of memory, then it swaps, but swap is on ZFS, so ZFS needs to allocate itself memory so it can do the work to allocate swap, so it needs to swap to free memory to do so....
bharatak: Right, you should leave space for the userland.
vm.kmem_size is supposed to limit kernel memory.  If you keep it below what you need for your userland, you should never run into swapping problems.
pilgrimd: I think i'm starting to grok it.. i'll make the tweaks and see.
so the prefetch thing, that just reads disk ahead of when its needed and stores it in the arc?
Right.  If you have lots of ARC, it can help performance by getting some blocks into the cache right before they're actually requested.
But if you don't have lots of ARC, all you really do is churn ARC.
got it
My home fileserver has 12 GB, with 6 GB ARC.  Usage swings between 3 GB and 5.4 GB, depending on how much we're streaming.
will make my setup worse
The ZFS mantra is always, "You can just install more RAM, right?" :)
i know, at my last job i had a setup with 24gb of ram, 24 disks
too bad the place is letting the system die, last i heard they wanted to put linux on it
thanks for the help!
already the iostat looks better ;l)
glad to be a customer of ARP networks again
shmget: :)
shmget: why did you leave b4? :)
NSD ftw
postgresql ftw... ;-)
cat, |, and grep  ftw
relational database schmatabase
MS DNS & MSSQL ftw :P
powerdns is wonderful
it's also much more mature than it was a few years ago
however, the cursor is still not threaded i don't think
PowerDNS will be for allowing customers to edit reverse DNS PTR's, then BIND or NSD slaves will suck the zones and be authoritative
should be fairly simple
that sounds good
yeah cuz PowerDNS with MySQL backend should integrate pretty straightforwardly into the Portal (a Rails app)
does anyone use sinatra and can suggest an easy and straightforward way to authenticate users?
homosaur: mount a rack middleware. devise, or rack-openid.
 hello, my vps seems to be unreachable.  when I boot it, I get a connection refused using vnc and [connecting....down] on the serial console
I figure I would at least get the bios post-like messages on vnc no matter what I've done to the vps install right?
hm.. ##mikrotik's topic ends with "If nobody answers right away, please wait. Most of us don't glue our eyes to IRC."
perhaps we could have something like that, maybe add to it about e-mailing support