Search results

I had a similar experience and whilst I didn't spend too much time on it I could reproduce the behaviour by flooding the NICs. In my case my storage is on a SAN and if I set off 10 'dd if=/dev/zero' jobs on the host then I could kiss goodbye to the server. Might be a red herring, but are you...

Maky sure you do: time 'dd if=/dev/zero of=/home/10G bs=1M count=10240; sync' Also, you need a minimum of 4 drives to do RAID10....

A couple of points ;) - that is so quick it has to be cached. Try using a file bigger than your VM RAM (i.e. 10GB or something) - secondly - the MBs doesn't include the sync. You want to do something like time 'dd if=/dev/zero of=/tmp/10G bs=1M count=10240; sync' (note the quotes)

Generally raw and virtio is the way to go. There are lots of posts about this, try searching (I don't mean that to sound as snarky as it does ;)).

Yep, *but* be careful - I did this originally (using shorewall) but every time I wanted to add a new public IP I had to restart the machine. This is less painful when it is a VM but very painful when it is the physical machine.

If you install it on a dedicated VM the idea is that every other VM hides behind it and the dedicated firewall VM does NAT/port forwarding to the 'internal' VMs. You certainly don't need to install it on every VM ;)

Thanks Udo! Now if you could do that with RAID10 and demonstrate how 4, 8 and 16 disks perform, I would be very appreciative :) :)

pveperf /var/lib/vz and pveperf / are useful diagnostics on the host also, iostat in the guest is quite useful....post those here.

A dedicated VM.

I think a lot of people (I do anyway ;)) run a software firewall (i.e. inside a VM) and then have that act as the gateway for all the other VMs across the cluster. So install something like vyatta or shorewall or pfsense install a VM, configure it to be a DHCP server and then install your other...

I have a three node cluster with 7 different VLANs. Each node has 4 NICs (2 for SAN, 2 for rest of the network) and I tried a number of them (pfsense, clarkos, ebox etc.). In the end I chose vyatta for a number of reasons: - command line auto complete is very convenient - entire...

Can you post your /etc/network/interfaces and also the output of ifconfig on each node

Assuming that the proxmox configuration is the same in terms of networking then yes, I think so.

No idea if your motherboard and chipset is supported but for the disks, 8x300 is better than 4x600. And yes, a decent BBU is absolutely critical. Also, stick some more RAM in there if you can - once you get bitten by this you will find loads more things to virtualise ;)

Ah - that explains it all. I had a similar situation with them - are you running hardware RAID? If so you need the battery backed RAID Controller cache unit (BBU). Your FSYNCs really need to be in the thousands as this is an indicator of the capability of your concurrent IO performance...

Not as far as I know, but I don't really use OpenVZ.

http://bitbud.com/2009/11/27/detailed-historical-performance-monitoring-for-proxmox/ looks interesting as well.

Wow - talk about a question with a million answers :) I personally use zabbix to monitor the servers and puppet to configure the servers. If you have a lot of servers then zabbix and puppet are very valuable. Nagios is really the granddaddy of them all and I think a lot can be learnt from...

It is impossible to answer this accurately as you haven't indicated the resource profile of the existing hardware servers. In my experience if you have lots and lots of machines that aren't IO bound then yes, sure, they make ideal candidates for VMs. *However*, even with virtio disk IO is...

I don't think so - I think the networking has to be done manually on each node. Not sure if you *can* use the GUI to change the network settings on nodes? I would do this by hand (unless someone trumps me with positive confirmation that the GUI will do the right thing)....