Recent content by aderumier

Before add it to gui, I think we need to add a deny with iothread and multiple disks, as it's buggy. also need to verify how it's works with zfs local replication. (maybe simply forbid it)

can you try without "numa: 1" ?

does it work for you ? (I don't have tested it)

@GPLExpert can you give me result of "cat /proc/cpuinfo" of a working node, and node2 ?

Ok, I have tested with iothread, and I have problem , migration crashing or qemu process crashing. Also with only 1disk. So it seem that qemu is buggy currently, for drive-mirror + livemigration at the same time when iothread is enabled. https://bugzilla.redhat.com/show_bug.cgi?id=1539530

yes, it should work. maybe a little bit overkill. you can give a try at https://www.openattic.org/ or wait for next ceph release (mimic), which should have integrated dashboards with management (create/delete/update).

you don't need to define vlan interfaces in /etc/network/interfaces if you define vlan tag in vm configuration, proxmox will create the bond0.vlan interface and a vmbr0v[vlan] bridge for you.

they are running fine (around same perf than 3610 or 3710), but take care of dwpd 1.

Hi, I'm currently working on implementation on vxlan + bgp evpn. This should give us something like vmware nsx. (with anycast gateway on proxmox host). This will work with linux bridge. I'll try to send patches next month.

it's only supported since win8/2012. (maybe 8.1,2012r2) (you also have a gui, "optimize disk")

note that since luminous + bluestore, jemalloc don't work well. (because of rocksdb) Ceph devs said that tcmalloc if fine now, since they have switched to async messenger.

if you are concerned about dataloss, cache=none. rbd_cache is 32mb (can be tuned), so even with fsync you can loose 32mb. (but you'll don't have filesystem corruption).

@David : do you have tried with a bigger file size ? (as it's random write, with a small file, you have more chance to have 2 block near each others, so writeback is usefull is this case).

if you enable cache=writeback on vm, it'll enable rbd_cache=true. ceph have a feature by default rbd cache writethrough until flush = true. That mean that it's waiting to receive a first fsync, before really enable writeback. So you are safe to enable writeback. Writeback is helping for...

The problem is coming from network latency + ceph latency. If you copy 1 file, sequentially and with small blocks, it's iodepth=1. (same with dd command for example). For each block, you'll have your network latency (0,1ms for example), you'll be able to do 10000 iops. if you do it with 4k...