Hello,
Having another issue here that I can't figure out.
I have a 2-node cluster here, running Proxmox 4.4.
This cluster is using shared storage that connects to a Linux storage
server using iSCSI via direct (NIC-to-NIC) 10-Gbit LAN (storage server
has 2x10-Gbit ports network cards; Proxmox nodes also have 10Gbit LAN NICs).
My storage model for the VMs is LVM-over-iSCSI (storage.cfg from one of
the nodes):
iscsi: Storage
portal 192.168.4.1
target iqn.2016-03.eu.company:storage.all.luns
content none
lvm: WWW-hosts
vgname vg-www-hosts
base
Storage:0.0.1.scsi-1494554000000000060fd8f3f2072a735bf2387ea56fe72a0
content images
shared 1
nfs: vmnfs
export /mnt/vmnfs
path /mnt/pve/vmnfs
server 192.168.4.1
content images
maxfiles 1
options vers=4
Problem is that LVM access to the virtual machine's LVs is terribly
slow, while NFS transfers are fast enough (storage server has 4x4TB SAS
drives in RAID10 configured with MDADM).
Here's 2 simple test I performed to get the transfer speeds:
1) For the LVM-over-iSCSI transfers:
(on the proxmox node)
# dd if=/dev/vg-www-hosts/vm-82105-disk-1 of=vm-82105-disk-1 bs=1M
count=3000
3000+0 records in
3000+0 records out
3145728000 bytes (3.1 GB) copied, 70.0969 s, 44.9 MB/s
#
2) For the NFS transfer from the same storage server (NFS is configured
as above):
root@pm1:~# dd if=/mnt/pve/ISOimages/template/iso/myOS.iso of=myOS.iso
bs=1M count=3000
3000+0 records in
3000+0 records out
3145728000 bytes (3.1 GB) copied, 1.17121 s, 2.7 GB/s
root@pm1:~#
So I cannot figure out why LVM-over-iSCSI is so slow. I'm using
iscsitarget package on Debian Jessie on the storage server as a iSCSI
target. Could this be due to iscsitarget package? As I wrote earlier,
iscsitarget appears to be abandoned in Debian Stretch, which leads me to
question if this is the right iSCSI target software I'm using here..
Thanks!
_______________________________________________
pve-user mailing list
[email protected]
https://pve.proxmox.com/cgi-bin/mailman/listinfo/pve-user
