I have recently installed Xen 3.0.3 from Debian binaries (Etch) which
was very painless and made the installation process very simple. I was
very impressed.
Having setup/configured Xen for 3 domUs, using xen-tools to setup the
machines, so far so good; that is until one of the guest domains
encounters a slight load.
Upon a domU encountering load (e.g. an rsync, make, ...) network
performance becomes poor, and ICMP works as proof of concept; whilst the
domU is under load, every ping consistently rises to 30-80ms - and this
is just pinging the dom0 IP from a domU! Note that I initially thought
this could have been a NIC issue as eth0 is a tg3-based NIC, however
swapping the cat5 over to eth1 (e100-based NIC) the problem still
exists. I am using bridging.
I have tried numerous things, including turning off TX checksumming
using ethtool and even tried changing the scheduler by adding
'sched=bvt' to my grub's conf; no change.
Looking at ifconfig, the vif interfaces show lots of dropped packets.
The issue, however, is not so much loss but general latency and slow
down (I, for example, notice SSH lag when a domU is under load).
However, in case it is related, a paste of a vif interface from
ifconfig:
vif1.0 Link encap:Ethernet HWaddr FE:FF:FF:FF:FF:FF
inet6 addr: fe80::fcff:ffff:feff:ffff/64 Scope:Link
UP BROADCAST RUNNING NOARP MTU:1500 Metric:1
RX packets:1065439 errors:0 dropped:0 overruns:0 frame:0
TX packets:1159440 errors:0 dropped:1819 overruns:0 carrier:0
collisions:0 txqueuelen:0
RX bytes:314895987 (300.3 MiB) TX bytes:145316240 (138.5 MiB)
I have no idea as to what to try next. Has anyone ever experienced these
issues before? I have been a user of Xen for a while now, having had
two domU VMs with two separate UK-based providers, and have never seen
these problems.
Many thanks.
Regards,
--
Simon
_______________________________________________
Xen-users mailing list
Xen-users@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-users
|
Home