Hey,
we've been running a Proxmox Cluster for years now, first a 1.9, later a 3.4. Those were/are Blade Center S
with Blades of 24 cores and 110GB Ram, running on a nfs storage.
Now the first machine with a 5.0-23 System, running on 8 cores and 32GB Ram - a prototype to test the
System before migrating. It only runs a single VM.
We've had this issue with a Windows 7 Client on the old system before: Insane amounts of pakets lost even
between host and server (talking 59% up). This VM is not used anymore and we did not have these issues
for a while. Now with the new system.
It runs a Windows 2012R2 Server on an local, inbuild storage, qcow2 format on ide1 with Default (No cache).
It has an additional disk on the same storage on sata0 and Default (No cache). 4 Cores on a single socket
without NUMA of the Defaul (kvm64) type - which run at a rather constant 25-30% load (the host runs at
12-17%). Automatically allocated memory between 1MB and 16384MB. Network is a Bridge to vmbr0 using
an Intel E1000 Type.
I've run ping tests from the Client Network (ClNet), the Server Network (SvNet) and back, getting the following:
Ping ClNet to Hos t : 0%
Ping ClNet to VM : 35%
Ping SvNet to Host : 0%
Ping SvNet to VM : 39%
Ping Host to VM : 1%
Ping VM to Host : 0%
Ping VM to ClNet: 9%
Ping VM to SvNet: 8%
The VM runs an heavily sql dependent application and receives very frequent timeouts which makes the
whole thing boarderline unusable. I tried switching the machine to virtio network drivers but this resulted
in the paket loss to climb beyond 50%.
Any ideas or tips or further questions to help with this welcome,
thank you in advance
we've been running a Proxmox Cluster for years now, first a 1.9, later a 3.4. Those were/are Blade Center S
with Blades of 24 cores and 110GB Ram, running on a nfs storage.
Now the first machine with a 5.0-23 System, running on 8 cores and 32GB Ram - a prototype to test the
System before migrating. It only runs a single VM.
We've had this issue with a Windows 7 Client on the old system before: Insane amounts of pakets lost even
between host and server (talking 59% up). This VM is not used anymore and we did not have these issues
for a while. Now with the new system.
It runs a Windows 2012R2 Server on an local, inbuild storage, qcow2 format on ide1 with Default (No cache).
It has an additional disk on the same storage on sata0 and Default (No cache). 4 Cores on a single socket
without NUMA of the Defaul (kvm64) type - which run at a rather constant 25-30% load (the host runs at
12-17%). Automatically allocated memory between 1MB and 16384MB. Network is a Bridge to vmbr0 using
an Intel E1000 Type.
I've run ping tests from the Client Network (ClNet), the Server Network (SvNet) and back, getting the following:
Ping ClNet to Hos t : 0%
Ping ClNet to VM : 35%
Ping SvNet to Host : 0%
Ping SvNet to VM : 39%
Ping Host to VM : 1%
Ping VM to Host : 0%
Ping VM to ClNet: 9%
Ping VM to SvNet: 8%
The VM runs an heavily sql dependent application and receives very frequent timeouts which makes the
whole thing boarderline unusable. I tried switching the machine to virtio network drivers but this resulted
in the paket loss to climb beyond 50%.
Any ideas or tips or further questions to help with this welcome,
thank you in advance