Re: [Openstack] [ceph-users] Openstack with Ceph, boot from volume
On 05/30/2013 02:50 PM, Martin Mailand wrote: Hi Josh, now everything is working, many thanks for your help, great work. Great! I added those settings to http://ceph.com/docs/master/rbd/rbd-openstack/ so it's easier to figure out in the future. -martin On 30.05.2013 23:24, Josh Durgin wrote: I have to more things. 1. The volume_driver=cinder.volume.driver.RBDDriver is deprecated, update your configuration to the new path. What is the new path? cinder.volume.drivers.rbd.RBDDriver 2. I have in the glance-api.conf show_image_direct_url=True, but the volumes are not clones of the original which are in the images pool. Set glance_api_version=2 in cinder.conf. The default was changed in Grizzly. ___ Mailing list: https://launchpad.net/~openstack Post to : openstack@lists.launchpad.net Unsubscribe : https://launchpad.net/~openstack More help : https://help.launchpad.net/ListHelp
Re: [Openstack] [ceph-users] Openstack with Ceph, boot from volume
Hi Josh, now everything is working, many thanks for your help, great work. -martin On 30.05.2013 23:24, Josh Durgin wrote: >> I have to more things. >> 1. The volume_driver=cinder.volume.driver.RBDDriver is deprecated, >> update your configuration to the new path. What is the new path? > > cinder.volume.drivers.rbd.RBDDriver > >> 2. I have in the glance-api.conf show_image_direct_url=True, but the >> volumes are not clones of the original which are in the images pool. > > Set glance_api_version=2 in cinder.conf. The default was changed in > Grizzly. ___ Mailing list: https://launchpad.net/~openstack Post to : openstack@lists.launchpad.net Unsubscribe : https://launchpad.net/~openstack More help : https://help.launchpad.net/ListHelp
Re: [Openstack] [ceph-users] Openstack with Ceph, boot from volume
On 05/30/2013 02:18 PM, Martin Mailand wrote: Hi Josh, that's working. I have to more things. 1. The volume_driver=cinder.volume.driver.RBDDriver is deprecated, update your configuration to the new path. What is the new path? cinder.volume.drivers.rbd.RBDDriver 2. I have in the glance-api.conf show_image_direct_url=True, but the volumes are not clones of the original which are in the images pool. Set glance_api_version=2 in cinder.conf. The default was changed in Grizzly. That's what I did. root@controller:~/vm_images# !1228 glance add name="Precise Server" is_public=true container_format=ovf disk_format=raw < ./precise-server-cloudimg-amd64-disk1.raw Added new image with ID: 6fbf4dfd-adce-470b-87fe-9b6ddb3993c8 root@controller:~/vm_images# rbd -p images -l ls NAMESIZE PARENT FMT PROT LOCK 6fbf4dfd-adce-470b-87fe-9b6ddb3993c8 2048M 2 6fbf4dfd-adce-470b-87fe-9b6ddb3993c8@snap 2048M 2 yes root@controller:~/vm_images# cinder create --image-id 6fbf4dfd-adce-470b-87fe-9b6ddb3993c8 --display-name volcli1 10+-+--+ | Property |Value | +-+--+ | attachments | [] | | availability_zone | nova | | bootable |false | | created_at | 2013-05-30T21:08:16.506094 | | display_description | None | | display_name| volcli1| | id | 34838911-6613-4140-93e0-e1565054a2d3 | | image_id | 6fbf4dfd-adce-470b-87fe-9b6ddb3993c8 | | metadata | {} | | size| 10 | | snapshot_id | None | | source_volid| None | |status | creating | | volume_type | None | +-+--+ root@controller:~/vm_images# cinder list +--+-+--+--+-+--+-+ | ID |Status | Display Name | Size | Volume Type | Bootable | Attached to | +--+-+--+--+-+--+-+ | 34838911-6613-4140-93e0-e1565054a2d3 | downloading | volcli1| 10 | None| false | | +--+-+--+--+-+--+-+ root@controller:~/vm_images# rbd -p volumes -l ls NAME SIZE PARENT FMT PROT LOCK volume-34838911-6613-4140-93e0-e1565054a2d3 10240M 2 root@controller:~/vm_images# -martin On 30.05.2013 22:56, Josh Durgin wrote: On 05/30/2013 01:50 PM, Martin Mailand wrote: Hi Josh, I found the problem, nova-compute tries to connect to the publicurl (xxx.xxx.240.10) of the keytone endpoints, this ip is not reachable from the management network. I thought the internalurl is the one, which is used for the internal communication of the openstack components and the publicurl is the ip for "customer" of the cluster? Am I wrong here? I'd expect that too, but it's determined in nova by the cinder_catalog_info option, which defaults to volume:cinder:publicURL. You can also override it explicitly with cinder_endpoint_template=http://192.168.192.2:8776/v1/$(tenant_id)s in your nova.conf. Josh -martin On 30.05.2013 22:22, Martin Mailand wrote: Hi Josh, On 30.05.2013 21:17, Josh Durgin wrote: It's trying to talk to the cinder api, and failing to connect at all. Perhaps there's a firewall preventing that on the compute host, or it's trying to use the wrong endpoint for cinder (check the keystone service and endpoint tables for the volume service). the keystone endpoint looks like this: | dd21ed74a9ac4744b2ea498609f0a86e | RegionOne | http://xxx.xxx.240.10:8776/v1/$(tenant_id)s | http://192.168.192.2:8776/v1/$(tenant_id)s | http://192.168.192.2:8776/v1/$(tenant_id)s | 5ad684c5a0154c13b54283b01744181b where 192.168.192.2 is the IP from the controller node. And from the compute node a telnet 192.168.192.2 8776 is working. -martin ___ ceph-users mailing list ceph-us...@lists.ceph.com http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com ___ Mailing list: https://launchpad.net/~openstack Post to : openstack@lists.launchpad.net Unsubscribe : https://launchpad.net/~openstack More help : https://help.launchpad.net/ListHelp
Re: [Openstack] [ceph-users] Openstack with Ceph, boot from volume
Hi Josh, that's working. I have to more things. 1. The volume_driver=cinder.volume.driver.RBDDriver is deprecated, update your configuration to the new path. What is the new path? 2. I have in the glance-api.conf show_image_direct_url=True, but the volumes are not clones of the original which are in the images pool. That's what I did. root@controller:~/vm_images# !1228 glance add name="Precise Server" is_public=true container_format=ovf disk_format=raw < ./precise-server-cloudimg-amd64-disk1.raw Added new image with ID: 6fbf4dfd-adce-470b-87fe-9b6ddb3993c8 root@controller:~/vm_images# rbd -p images -l ls NAMESIZE PARENT FMT PROT LOCK 6fbf4dfd-adce-470b-87fe-9b6ddb3993c8 2048M 2 6fbf4dfd-adce-470b-87fe-9b6ddb3993c8@snap 2048M 2 yes root@controller:~/vm_images# cinder create --image-id 6fbf4dfd-adce-470b-87fe-9b6ddb3993c8 --display-name volcli1 10+-+--+ | Property |Value | +-+--+ | attachments | [] | | availability_zone | nova | | bootable |false | | created_at | 2013-05-30T21:08:16.506094 | | display_description | None | | display_name| volcli1| | id | 34838911-6613-4140-93e0-e1565054a2d3 | | image_id | 6fbf4dfd-adce-470b-87fe-9b6ddb3993c8 | | metadata | {} | | size| 10 | | snapshot_id | None | | source_volid| None | |status | creating | | volume_type | None | +-+--+ root@controller:~/vm_images# cinder list +--+-+--+--+-+--+-+ | ID |Status | Display Name | Size | Volume Type | Bootable | Attached to | +--+-+--+--+-+--+-+ | 34838911-6613-4140-93e0-e1565054a2d3 | downloading | volcli1| 10 | None| false | | +--+-+--+--+-+--+-+ root@controller:~/vm_images# rbd -p volumes -l ls NAME SIZE PARENT FMT PROT LOCK volume-34838911-6613-4140-93e0-e1565054a2d3 10240M 2 root@controller:~/vm_images# -martin On 30.05.2013 22:56, Josh Durgin wrote: > On 05/30/2013 01:50 PM, Martin Mailand wrote: >> Hi Josh, >> >> I found the problem, nova-compute tries to connect to the publicurl >> (xxx.xxx.240.10) of the keytone endpoints, this ip is not reachable from >> the management network. >> I thought the internalurl is the one, which is used for the internal >> communication of the openstack components and the publicurl is the ip >> for "customer" of the cluster? >> Am I wrong here? > > I'd expect that too, but it's determined in nova by the > cinder_catalog_info option, which defaults to volume:cinder:publicURL. > > You can also override it explicitly with > cinder_endpoint_template=http://192.168.192.2:8776/v1/$(tenant_id)s > in your nova.conf. > > Josh > >> -martin >> >> On 30.05.2013 22:22, Martin Mailand wrote: >>> Hi Josh, >>> >>> On 30.05.2013 21:17, Josh Durgin wrote: It's trying to talk to the cinder api, and failing to connect at all. Perhaps there's a firewall preventing that on the compute host, or it's trying to use the wrong endpoint for cinder (check the keystone service and endpoint tables for the volume service). >>> >>> the keystone endpoint looks like this: >>> >>> | dd21ed74a9ac4744b2ea498609f0a86e | RegionOne | >>> http://xxx.xxx.240.10:8776/v1/$(tenant_id)s | >>> http://192.168.192.2:8776/v1/$(tenant_id)s | >>> http://192.168.192.2:8776/v1/$(tenant_id)s | >>> 5ad684c5a0154c13b54283b01744181b >>> >>> where 192.168.192.2 is the IP from the controller node. >>> >>> And from the compute node a telnet 192.168.192.2 8776 is working. >>> >>> -martin >>> ___ >>> ceph-users mailing list >>> ceph-us...@lists.ceph.com >>> http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com >>> > ___ Mailing list: https://launchpad.net/~openstack Post to : openstack@lists.launchpad.net Unsubscribe : https://launchpad.net/~openstack More help : https://help.launchpad.net/ListHelp
Re: [Openstack] [ceph-users] Openstack with Ceph, boot from volume
On 05/30/2013 01:50 PM, Martin Mailand wrote: Hi Josh, I found the problem, nova-compute tries to connect to the publicurl (xxx.xxx.240.10) of the keytone endpoints, this ip is not reachable from the management network. I thought the internalurl is the one, which is used for the internal communication of the openstack components and the publicurl is the ip for "customer" of the cluster? Am I wrong here? I'd expect that too, but it's determined in nova by the cinder_catalog_info option, which defaults to volume:cinder:publicURL. You can also override it explicitly with cinder_endpoint_template=http://192.168.192.2:8776/v1/$(tenant_id)s in your nova.conf. Josh -martin On 30.05.2013 22:22, Martin Mailand wrote: Hi Josh, On 30.05.2013 21:17, Josh Durgin wrote: It's trying to talk to the cinder api, and failing to connect at all. Perhaps there's a firewall preventing that on the compute host, or it's trying to use the wrong endpoint for cinder (check the keystone service and endpoint tables for the volume service). the keystone endpoint looks like this: | dd21ed74a9ac4744b2ea498609f0a86e | RegionOne | http://xxx.xxx.240.10:8776/v1/$(tenant_id)s | http://192.168.192.2:8776/v1/$(tenant_id)s | http://192.168.192.2:8776/v1/$(tenant_id)s | 5ad684c5a0154c13b54283b01744181b where 192.168.192.2 is the IP from the controller node. And from the compute node a telnet 192.168.192.2 8776 is working. -martin ___ ceph-users mailing list ceph-us...@lists.ceph.com http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com ___ Mailing list: https://launchpad.net/~openstack Post to : openstack@lists.launchpad.net Unsubscribe : https://launchpad.net/~openstack More help : https://help.launchpad.net/ListHelp
Re: [Openstack] [ceph-users] Openstack with Ceph, boot from volume
Hi Josh, I found the problem, nova-compute tries to connect to the publicurl (xxx.xxx.240.10) of the keytone endpoints, this ip is not reachable from the management network. I thought the internalurl is the one, which is used for the internal communication of the openstack components and the publicurl is the ip for "customer" of the cluster? Am I wrong here? -martin On 30.05.2013 22:22, Martin Mailand wrote: > Hi Josh, > > On 30.05.2013 21:17, Josh Durgin wrote: >> It's trying to talk to the cinder api, and failing to connect at all. >> Perhaps there's a firewall preventing that on the compute host, or >> it's trying to use the wrong endpoint for cinder (check the keystone >> service and endpoint tables for the volume service). > > the keystone endpoint looks like this: > > | dd21ed74a9ac4744b2ea498609f0a86e | RegionOne | > http://xxx.xxx.240.10:8776/v1/$(tenant_id)s | > http://192.168.192.2:8776/v1/$(tenant_id)s | > http://192.168.192.2:8776/v1/$(tenant_id)s | > 5ad684c5a0154c13b54283b01744181b > > where 192.168.192.2 is the IP from the controller node. > > And from the compute node a telnet 192.168.192.2 8776 is working. > > -martin > ___ > ceph-users mailing list > ceph-us...@lists.ceph.com > http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com > ___ Mailing list: https://launchpad.net/~openstack Post to : openstack@lists.launchpad.net Unsubscribe : https://launchpad.net/~openstack More help : https://help.launchpad.net/ListHelp
Re: [Openstack] [ceph-users] Openstack with Ceph, boot from volume
Hi Weiguo, my answers are inline. -martin On 30.05.2013 21:20, w sun wrote: > I would suggest on nova compute host (particularly if you have > separate compute nodes), > > (1) make sure "rbd ls -l -p " works and /etc/ceph/ceph.conf is > readable by user nova!! yes to both > (2) make sure you can start up a regular ephemeral instance on the > same nova node (ie, nova-compute is working correctly) an ephemeral instance is working > (3) if you are using cephx, make sure libvirt secret is set up correct > per instruction at ceph.com I do not use cephx > (4) look at /var/lib/nova/instance/x/libvirt.xml and the > disk file is pointing to the rbd volume For an ephemeral instance the folder is create, for a volume bases instance the folder is not created. > (5) If all above look fine and you still couldn't perform nova boot > with the volume, you can try last thing to manually start up a kvm > session with the volume similar to below. At least this will tell you > if you qemu has the correct rbd enablement. > > /usr/bin/kvm -m 2048 -drive > file=rbd:ceph-openstack-volumes/volume-3f964f79-febe-4251-b2ba-ac9423af419f,index=0,if=none,id=drive-virtio-disk0 > -boot c -net nic -net user -nographic -vnc :1000 -device > piix3-usb-uhci,id=usb,bus=pci.0,addr=0x1.0x2 -device > virtio-blk-pci,scsi=off,bus=pci.0,addr=0x4,drive=drive-virtio-disk0,id=virtio-disk0,bootindex=1 > If I start kvm by hand it is working. > --weiguo > > > Date: Thu, 30 May 2013 16:37:40 +0200 > > From: mar...@tuxadero.com > > To: ceph-us...@ceph.com > > CC: openstack@lists.launchpad.net > > Subject: [ceph-users] Openstack with Ceph, boot from volume > > > > Hi Josh, > > > > I am trying to use ceph with openstack (grizzly), I have a multi > host setup. > > I followed the instruction > http://ceph.com/docs/master/rbd/rbd-openstack/. > > Glance is working without a problem. > > With cinder I can create and delete volumes without a problem. > > > > But I cannot boot from volumes. > > I doesn't matter if use horizon or the cli, the vm goes to the error > state. > > > > From the nova-compute.log I get this. > > > > 2013-05-30 16:08:45.224 ERROR nova.compute.manager > > [req-5679ddfe-79e3-4adb-b220-915f4a38b532 > > 8f9630095810427d865bc90c5ea04d35 43b2bbbf5daf4badb15d67d87ed2f3dc] > > [instance: 059589a3-72fc-444d-b1f0-ab1567c725fc] Instance failed block > > device setup > > . > > 2013-05-30 16:08:45.224 19614 TRACE nova.compute.manager [instance: > > 059589a3-72fc-444d-b1f0-ab1567c725fc] ConnectionError: [Errno 101] > > ENETUNREACH > > > > What tries nova to reach? How could I debug that further? > > > > Full Log included. > > > > -martin > > > > Log: > > > > ceph --version > > ceph version 0.61 (237f3f1e8d8c3b85666529860285dcdffdeda4c5) > > > > root@compute1:~# dpkg -l|grep -e ceph-common -e cinder > > ii ceph-common 0.61-1precise > > common utilities to mount and interact with a ceph storage > > cluster > > ii python-cinderclient 1:1.0.3-0ubuntu1~cloud0 > > python bindings to the OpenStack Volume API > > > > > > nova-compute.log > > > > 2013-05-30 16:08:45.224 ERROR nova.compute.manager > > [req-5679ddfe-79e3-4adb-b220-915f4a38b532 > > 8f9630095810427d865bc90c5ea04d35 43b2bbbf5daf4badb15d67d87ed2f3dc] > > [instance: 059589a3-72fc-444d-b1f0-ab1567c725fc] Instance failed block > > device setup > > 2013-05-30 16:08:45.224 19614 TRACE nova.compute.manager [instance: > > 059589a3-72fc-444d-b1f0-ab1567c725fc] Traceback (most recent call last): > > 2013-05-30 16:08:45.224 19614 TRACE nova.compute.manager [instance: > > 059589a3-72fc-444d-b1f0-ab1567c725fc] File > > "/usr/lib/python2.7/dist-packages/nova/compute/manager.py", line 1071, > > in _prep_block_device > > 2013-05-30 16:08:45.224 19614 TRACE nova.compute.manager [instance: > > 059589a3-72fc-444d-b1f0-ab1567c725fc] return > > self._setup_block_device_mapping(context, instance, bdms) > > 2013-05-30 16:08:45.224 19614 TRACE nova.compute.manager [instance: > > 059589a3-72fc-444d-b1f0-ab1567c725fc] File > > "/usr/lib/python2.7/dist-packages/nova/compute/manager.py", line 721, in > > _setup_block_device_mapping > > 2013-05-30 16:08:45.224 19614 TRACE nova.compute.manager [instance: > > 059589a3-72fc-444d-b1f0-ab1567c725fc] volume = > > self.volume_api.get(context, bdm['volume_id']) > > 2013-05-30 16:08:45.224 19614 TRACE nova.compute.manager [instance: > > 059589a3-72fc-444d-b1f0-ab1567c725fc] File > > "/usr/lib/python2.7/dist-packages/nova/volume/cinder.py", line 193, > in get > > 2013-05-30 16:08:45.224 19614 TRACE nova.compute.manager [instance: > > 059589a3-72fc-444d-b1f0-ab1567c725fc] > > self._reraise_translated_volume_exception(volume_id) > > 2013-05-30 16:08:45.224 19614 TRACE nova.compute.manager [instance: > > 059589a3-72fc-444d-b1f0-ab1567c725fc] File > > "/usr/lib/python2.7/dist-packages/nova/volume/cinder.py", line 190, > in get > > 2013-05-30 16:08:45.224 19614 TRACE nova.compute.manager [instance: > > 059589a3-72fc-444d-b1f0-ab1567c725fc] item = > > cinderclien