System hangs when using NFS storage backend with loopback mounts
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
OpenStack Compute (nova) |
In Progress
|
Low
|
Unassigned |
Bug Description
Description
===========
When using high speed disks and NFS as storage backend, during high loads the nfs mounts hang indefinitely.
Steps to reproduce
==================
A chronological list of steps which will bring off the
issue you noticed:
* Spin up a VM with a mounted cinder volume from an NFS backend
* Generate some read/write load
* Occationally the loopback NFS mounts will hang. The machine and everything else using that mount will also hang.
Expected result
===============
The system should run stably
Actual result
=============
Occasionally , usually during higher load the system will hang.
Environment
===========
1. Exact version of OpenStack you are running. See the following
Openstack Kilo
openstack-
openstack-
python-
openstack-
openstack-
openstack-
python-
openstack-
openstack-
openstack-
2. Which hypervisor did you use?
Libvirt + KVM,
2. Which storage type did you use?
NFS
3. Which networking type did you use?
Neutron with OpenVSwitch
Logs & Configs
==============
Nova.conf:
[DEFAULT]
notification_
notification_
notification_driver =
notification_
rpc_backend=rabbit
internal_
default_
notify_
state_path=
report_interval=10
enabled_
ec2_listen=0.0.0.0
ec2_workers=2
osapi_compute_
osapi_compute_
metadata_
metadata_workers=2
compute_
service_
rootwrap_
auth_strategy=
use_forwarded_
novncproxy_
novncproxy_
allow_resize_
block_device_
heal_instance_
reserved_
network_
default_
force_snat_
metadata_
dhcp_domain=
security_
debug=True
verbose=True
log_dir=
use_syslog=False
cpu_allocation_
ram_allocation_
scheduler_
scheduler_
compute_
vif_plugging_
vif_plugging_
firewall_
remove_
force_raw_
novncproxy_
vncserver_
vncserver_
vnc_enabled=True
vnc_keymap=en-us
volume_
amqp_durable_
sql_connection=
lock_path=
osapi_volume_
[api_database]
[barbican]
[cells]
[cinder]
[conductor]
workers=2
[database]
[ephemeral_
[glance]
api_servers=
[guestfs]
[hyperv]
[image_file_url]
[ironic]
[keymgr]
[keystone_
auth_uri=http://
identity_uri=http://
admin_user=nova
admin_password=
[libvirt]
virt_type=kvm
inject_
inject_key=False
inject_partition=-1
live_migration_
cpu_mode=host-model
disk_cachemodes
nfs_mount_
vif_driver=
[metrics]
[neutron]
......
Cinder.conf:
[nfs_ssd]
nfs_used_ratio=0.95
nfs_oversub_
volume_
nfs_shares_
volume_
quota_volumes = -1
nfs_mount_
- No notable output in nova log
- System log /dmesg after a hang:
Nov 24 04:10:41 openstack1.
Nov 24 04:10:41 openstack1.
Nov 24 04:10:41 openstack1.
Nov 24 04:10:41 openstack1.
Nov 24 04:10:41 openstack1.
Nov 24 04:10:41 openstack1.
Nov 24 04:10:41 openstack1.
Nov 24 04:10:41 openstack1.
Nov 24 04:10:41 openstack1.
Nov 24 04:10:41 openstack1.
Nov 24 04:10:41 openstack1.
Nov 24 04:10:41 openstack1.
Nov 24 04:10:41 openstack1.
Nov 24 04:10:41 openstack1.
Nov 24 04:10:41 openstack1.
Nov 24 04:10:41 openstack1.
Nov 24 04:10:41 openstack1.
Nov 24 04:10:41 openstack1.
Nov 24 04:10:41 openstack1.
Nov 24 04:10:41 openstack1.
Nov 24 04:10:41 openstack1.
Nov 24 04:10:41 openstack1.
Nov 24 04:10:41 openstack1.
Changed in nova: | |
status: | Confirmed → In Progress |
I will post how we mitigate the problem when I formulate it.