Hi,
When testing on 3.8.1 I found, that shutdown/cancel/delete of VM in a
host is done in parallel without ability to configure concurrency (like
MAX_HOST in sched.conf). That leads to several problems:
(1) potential overload of storage with persistent images (ssh/qcow2 TM
drivers)
(2) race conditions in ebtables/iptables, as those tools doesn't support
concurrent execution
I saw (2) in my crash tests (submitting many VMs to the host with
network=ebtables and doing onevm shutdown on them), which could be
solved with "flock /tmp/ebtables sudo ebtables" and "flock /tmp/iptables
sudo iptables" in OpenNebulaNetwork.rb as a workaround.
When MAX_HOST=1 in sched.conf, then there are no problems with
ebtables/iptables concurrency in deploy time, but when MAX_HOST>1, then
the (2) problem could also be.
Regards, Rolandas Naujikas
_______________________________________________
Users mailing list
Users@lists.opennebula.org
http://lists.opennebula.org/listinfo.cgi/users-opennebula.org
