11,012 reputation
21536
bio website linkedin.com/in/dyasny
location Canada
age 34
visits member for 5 years, 3 months
seen 1 hour ago

Open source virt geek


4h
comment What should an Ops person know about hardware for an interview?
usually drives that get kicked out of an array, while remaining usable have bad firmware (or the controller does). bad blocks get relocated until there is no more room to relocate, and then the drive becomes physically faulty. bad firmware often causes scsi timeouts, which in turn cause the disk to fail to respond in an array, and get marked as faulty
7h
comment What should an Ops person know about hardware for an interview?
I was walking about TWO failed disks, not one. The idea of retagging is to bring the array up with only one disk missing, which is tolerable. The main trick is to know which of the two failed disks to keep
12h
comment What should an Ops person know about hardware for an interview?
It's a well known procedure called retagging. With two failed disks, you can't backup the data, but you can look at the logs, find out which drive failed first, and build a new array, focing the first failed disk offline. Always a good chance to get accessible data in a degraded array - good enough to slavage what needs to be salvaged
2d
comment What should an Ops person know about hardware for an interview?
ask them what they would do with a raid 5 in which two drives failed, within a day of each other. Both drives are still usable, and will not be replaced
Oct
20
comment Infiniband + KVM
I know for certain IPoIB setups with KVM work quite well, Mellanox have showcased quite a few use cases. IIRC, most of those were simple L2 bridges over an IB interface, with virtio_net in the guest.
Oct
11
comment Are /dev/disk/by-id symlinks unreliable?
if you're provisioning in a VM (is this TripleO?) the disk IDs may not be reliable, but under KVM the disks always show up in the sam eorder they are defined in the libvirt domxml. On real hardware, these IDs should be reliable afaik.
Oct
10
comment Redis vs Memcache
@CarlZulauf thanks ok, like I said, it's just customary to provide a short one liner disclaimer before singing such praises.
Oct
9
comment Redis vs Memcache
I don't intend to contest what you wrote, but... This sounds awfully like marketing. If you work for Pivotal or another company that specializes on Redis, it is common courtesy to provide a disclaimer before writing such a post. If not, you might as well provide a short disclaimer saying you don't. Otherwise, naturally, bias is suspect.
Oct
8
comment ssh into virtual machine (or direct VM console access)
Well, do automatic deployments with all the console options in the config scripts or the template and you will not have these issues. As for grub - the console option there is just to be able to see the server boot, after that, you need the ttyS0 setup in the guest as well
Oct
7
comment ssh into virtual machine (or direct VM console access)
if you're properly configured, hit enter at this point a couple of times
Oct
6
comment kvm error: Cannot find suitable CPU model for given data
can you post the output or virsh capabilities?
Oct
5
comment How To Monitor And Alert For KVM Hypervisors
it's simple. vmware created a system where hardware monitoring was impossible using standard tools, so they had to fulfill the technical debt themselves. KVM doesn't have this problem, you can monitor KVM hosts exactly the same way you would monitor any other host. RHEV is a virtualization management platform, it focuses on virtualization issues, and does it well. Hardware monitoring is done best by vendor tools, and you can use those directly, without resorting to 3rd party implementations
Oct
5
comment How To Monitor And Alert For KVM Hypervisors
@MichaelHampton monitorix, nagios, zenoss, pandoraFMS, munin and probably quite a few others have libvirt checks
Oct
5
comment How To Monitor And Alert For KVM Hypervisors
and to add to that, if you also install RHEV-Reports (it's optional, and comes free with RHEV), you get a history database running, and a reporting engine, so you can watch load, availability and other trends and history over time, troubleshoot those trends etc. There's also a free ManageIQ and VMTrubo appliances available for RHEV, also designed to deepdive into load trends and other criteria. Lots of absolutely free and serious BI functionality available there
Oct
5
comment How To Monitor And Alert For KVM Hypervisors
There's a large set of VM and other entity related events you can set up alerts and notifications for. And the list is of course extensible - statuses are visible through the API so a monitoring solution can check them
Oct
4
comment How To Monitor And Alert For KVM Hypervisors
VM states are tracked, and acted upon (for HA and load balancing policies), what do you mean by performance thresholds?
Oct
4
comment Virtualization-(Cluster) with KVM/libvirt on top of a FC SAN
two hosts aren't a cloud, I'd rule out all the cloud oriented software from the very start. You are trying to build something along the same lines as alteeve.ca/w/AN!Cluster_Tutorial_2, though if the system is going to grow, I'd opt for something more serious, like oVirt - it has all the features you asked for and much much more.
Oct
3
comment windows server 2012 4 processors
please see edit
Oct
3
comment How do I enable KVM device passthrough in Linux?
have you done a full powercycle after enabling VT and VT-d?
Oct
1
comment Access pty login prompt in VM
Well, I've seen plenty of cases when kvm/libvirt/qemu/etc misbehaved on ubuntu and worked perfectly on fedora, no surprises there :)