search for: 2node

Hi, i have a 2node ctdb-samba-cluster on debian squeeze-machines. Samba: 3.5.6 Ctdb: 1.0.112-12-1 Filesystem: glusterfs Sometimes i have the problem that some Office-Files are locked for other MS-Office-Users. If i try to access those locked files using linux(mount cifs and open it with openoffice) i can open it. I...

...cks and batch size of 1 = ticketlock). V2 implementation has been optimized for baremetal : (1) Making !CONFIG_PARAVIRT_SPINLOCK case equivalent to older ticketlock. (2) Removing extra overhead of cmpxchg for CONFIG_PARAVIRT_SPINLOCK case by replacing with add_smp call. Result: Test system: 32cpu 2node machine w/ 64GB each (16 pcpu machine +ht). Guests: 8GB 16vcpu guests (average of 8 iterations) A: batched ticket with batch_size = 4 B: batched ticket with batch_size = 8 C: qspinlock V11 of Waiman D: qspinlock V11 of Waiman with VIRT_UNFAIR=y % Improvements with kvm guests A B...

...cks and batch size of 1 = ticketlock). V2 implementation has been optimized for baremetal : (1) Making !CONFIG_PARAVIRT_SPINLOCK case equivalent to older ticketlock. (2) Removing extra overhead of cmpxchg for CONFIG_PARAVIRT_SPINLOCK case by replacing with add_smp call. Result: Test system: 32cpu 2node machine w/ 64GB each (16 pcpu machine +ht). Guests: 8GB 16vcpu guests (average of 8 iterations) A: batched ticket with batch_size = 4 B: batched ticket with batch_size = 8 C: qspinlock V11 of Waiman D: qspinlock V11 of Waiman with VIRT_UNFAIR=y % Improvements with kvm guests A B...

...er can not acquire more than batch_size times repeatedly during contention. On the negetive side we would need an extra cmpxchg. The patch has the batch size of 4. (As we know increasing batch size means we are closer to unfair locks and batch size of 1 = ticketlock). Result: Test system: 32cpu 2node machine w/ 64GB each (16 pcpu machine +ht). Guests: 8GB 16vcpu guests (average of 8 iterations) % Improvements with kvm guests (batch size = 4): ebizzy_0.5x 4.3 ebizzy_1.0x 7.8 ebizzy_1.5x 23.4 ebizzy_2.0x 48.6 Baremetal: ebizzy showed very high stdev, kernbench result was good b...

...er can not acquire more than batch_size times repeatedly during contention. On the negetive side we would need an extra cmpxchg. The patch has the batch size of 4. (As we know increasing batch size means we are closer to unfair locks and batch size of 1 = ticketlock). Result: Test system: 32cpu 2node machine w/ 64GB each (16 pcpu machine +ht). Guests: 8GB 16vcpu guests (average of 8 iterations) % Improvements with kvm guests (batch size = 4): ebizzy_0.5x 4.3 ebizzy_1.0x 7.8 ebizzy_1.5x 23.4 ebizzy_2.0x 48.6 Baremetal: ebizzy showed very high stdev, kernbench result was good b...

Hi Gilberto, hi all, thanks a lot for all your answers. At first I changed both settings mentioned below and first test look good. Before changing the settings I was able to crash a new installed VM every time after a fresh installation by producing much i/o, e.g. when installing Libre Office. This always resulted in corrupt files inside the VM, but researching the qcow2 file with the