10,690 reputation
21533
bio website linkedin.com/in/dyasny
location Canada
age 34
visits member for 5 years, 1 month
seen 5 hours ago

Open source virt geek


1d
comment Monitoring / metric collection for system collectives that change a lot in time (a.k.a. cloud)
please see edit
1d
comment Blade server backplane wiring
There aren't too many companies that produce blades, just call them, and ask to talk to a tech sales rep.
Aug
16
comment Poor network performance with KVM (virtio drivers) - Update: with vhost_net
I'd love to see the tweaks you have applied to get to over 40% in a blogpost or a KB article somewhere, might be useful
Aug
15
comment Can a Sun rack be converted to a normal rack?
Yup, Dell call those "versa rails", they tend to be slightly more expensive, but they are so much easier to work with
Aug
15
comment Can you use CGroups inside an OpenVZ Container?
Guess I was wrong, that can happen sometimes :)
Aug
15
comment Can you use CGroups inside an OpenVZ Container?
ouch, did I have a misconception there? I was sure in 7.0 it's still TP
Aug
15
comment Poor network performance with KVM (virtio drivers) - Update: with vhost_net
check the dell site, there might be newer firmware available for the broadcoms. My stack is almost entirely Dell as well (a bunch of R610 and R620) and I've no problems with performance without any tweaks
Aug
15
comment Can you use CGroups inside an OpenVZ Container?
I prefer to use LXC myself these days, but my use case and the typical use case for OVZ are quite different. I don't do hosting after all :) LXC has its issues, and my criteria for saying it is prod ready is when RHEL 7 officially supports LXC, sans the "tech preview"
Aug
15
comment Can you use CGroups inside an OpenVZ Container?
still, OVZ is the currently mature and stable tech, and it has it's own tools for resource management, that were implemented before cgroups ever came out. Not to downplay LXC, just trying to be fair
Aug
14
comment Poor network performance with KVM (virtio drivers) - Update: with vhost_net
on a gigabit link, usually about 60-70 mbytes/sec
Aug
14
comment Poor network performance with KVM (virtio drivers) - Update: with vhost_net
the ethX adapter under the bridge on the host actually
Aug
13
comment Poor network performance with KVM (virtio drivers) - Update: with vhost_net
ok, I'd start with ethtool -k and start playing with disabling TSO, LRO, GSO
Aug
13
comment Migrating Hyper-V VM's to CentOS 6
I don't really care about downvotes, but if anyone decides to do that, I expect a comment with an explanation.
Aug
13
comment What does cpu-pinning mean in context of KVM hypervisor with multiple virtual machines
I've seen performance drop on moderately loaded systems, when pinning was in use
Aug
13
comment Poor network performance with KVM (virtio drivers) - Update: with vhost_net
you mentioned jumbo frames, are they set up on the entire network stack and in the VMs?
Aug
7
comment KVM unix socket file permission for VNC
Try Fedora or CentOS, that's where libvirt/kvm/qemu are mainly developed
Aug
7
comment KVM unix socket file permission for VNC
And setting up the unix_sock_rw_perms param doesn't help? Sounds like a bug to me. Try another distro maybe...
Aug
7
comment Apache web server, virtual hosts, sub domains and multiple web frameworks
Zimbra is not likely to work well along other web apps, it is usually recommended to set it up separately
Aug
6
comment KVM unix socket file permission for VNC
Among other things, this config file controls the user libvirt generally runs under, which should also work for you. In any case, you are trying to override defaults that were determined to be optimal, especially for security reasons, I wouldn't change them, better to add your user to the allowed group instead
Aug
5
comment Different virtualization technologies without VT-d (IMMOU)
If you don't have the right hardware, just don't use virtualization. That's all there is to it. You can't have proper PCI passthrough that will work, be stable and perform without specific requirements, and if you can't satisfy those, just don't. It's very simple.