Perl 6 - the future is here, just unevenly distributed

IRC log for #fuel, 2014-03-14

| Channels | #fuel index | Today | | Search | Google Search | Plain-Text | summary

All times shown according to UTC.

Time Nick Message
00:25 alexz__ joined #fuel
00:54 bogdando joined #fuel
00:56 xarses joined #fuel
01:06 IlyaE joined #fuel
01:08 geogdog joined #fuel
01:17 xarses_ joined #fuel
01:18 geogdog joined #fuel
01:24 rmoe joined #fuel
01:26 vkozhukalov_ joined #fuel
02:04 justif well fail ubuntu and cinder deploy failed
02:18 dhblaz joined #fuel
02:30 dhblaz joined #fuel
02:40 ToTsiroll joined #fuel
02:45 designated joined #fuel
02:47 ilbot3 joined #fuel
02:47 Topic for #fuel is now Fuel 4.1 for Openstack: http://fuel.mirantis.com/ | Paste here http://paste.openstack.org/ | IRC logs http://irclog.perlgeek.de/fuel/
02:57 isAAAc joined #fuel
03:51 dhblaz joined #fuel
03:54 designated fresh fuel 4.1 install, trying to deploy HA, neutron/VLANs, ubuntu 12.04 with KVM.  Gets through ubuntu install and continues to fail when deploying openstack, same place everytime.
03:54 designated (/Stage[main]/Nova::Scheduler/Nova::Generic_service[scheduler]/Package[nova-scheduler]/ensure) change from purged to latest failed: Could not update: Execution of '/usr/bin/apt-get -q -y -o DPkg::Options::=--force-confold install nova-scheduler' returned 100: Reading package lists...
03:57 designated sorry it starts by failing with guestmount
03:57 designated (/Stage[main]/Nova::Utilities/Package[guestmount]/ensure) change from purged to present failed: Execution of '/usr/bin/apt-get -q -y -o DPkg::Options::=--force-confold install guestmount' returned 100: Reading package lists...
03:57 designated everything afterwards fails
04:03 fandi joined #fuel
04:14 dhblaz joined #fuel
05:25 Ch00k joined #fuel
05:26 IlyaE joined #fuel
05:57 dburmistrov joined #fuel
06:44 ToTsiroll hi..anyone tried importing ubuntu 12.04 precise cloud image ARM to openstack?
06:44 ToTsiroll i tried the release of ubuntu but it stops on mounting the disk
06:53 e0ne joined #fuel
07:11 vvalyavskiy joined #fuel
07:27 vvalyavskiy joined #fuel
07:45 vvalyavskiy Hello guys! I created cluster(simple mode), added two nodes to it, provisioned this nodes, made some modifications in deployment facts(.../deployment_id/co..._8.yaml ). Now, I want to get this yaml files, move it to relevant nodes and to start deployment locally using "puppet apply ....". Is fuel support this deployment way? - fuel-4.1
07:47 Ch00k joined #fuel
08:12 topochan joined #fuel
08:13 danirius joined #fuel
08:15 fandi joined #fuel
08:33 Ch00k joined #fuel
08:34 evgeniyl` vvalyavskiy: you can do it manually, first, copy puppet manifests from master node (we run rsync via mcollective to sync manifests), then copy yaml files to /etc/astute.yaml and I think you will be able to run puppet apply.
09:00 vvalyavskiy when I try to do it  on compute I get i following resut:   Parameter private_key_path failed on Install_ssh_keys[nova_ssh_key_for_migration]: /var/lib/astute/nova/nova no such file
09:12 tatyana joined #fuel
09:28 vvalyavskiy question is not actual now
09:33 Ch00k joined #fuel
09:37 rvyalov joined #fuel
09:41 glycerine joined #fuel
10:07 saju_m joined #fuel
10:09 Ch00k joined #fuel
10:23 DaveJ__ joined #fuel
10:27 Ch00k joined #fuel
10:34 tatyana joined #fuel
11:19 tatyana joined #fuel
11:24 TVR___ joined #fuel
11:29 rvyalov joined #fuel
11:50 Ch00k joined #fuel
11:52 TVR___ good morning all
11:56 dhblaz joined #fuel
11:57 e0ne joined #fuel
12:14 saju_m joined #fuel
12:15 dburmistrov joined #fuel
12:17 saju_m joined #fuel
12:22 justif joined #fuel
12:26 toha1 joined #fuel
12:42 warpig afternoon guys....
12:45 warpig anyone run into issues with radosgw deployment on 4.1?
12:45 warpig works fine on 4.0
12:45 warpig but there is no "containers" link in Horizon when deployed with Fuel 4.1
12:46 warpig "swift list" gives "Endpoint for object-store not found - have you specified a region?"
12:46 warpig after sourcing openrc
12:52 warpig CentOS deployment, btw...  :o)
13:05 baboune joined #fuel
13:06 baboune hello, got a question on a creation of bridges in the controller node.  We basiclly seem to have a bridge on every eth in the machine even though one (eth1) is not assigned.  I was wondering how this is done?  And how to assigne instead a static ip to this eth1?
13:06 baboune so in the UI, nothing is assigned to this eth card.  Yet a bridge is still created.
13:07 baboune wondering how and where?
13:08 baboune this is using Fuel 4.1
13:09 akasatkin joined #fuel
13:17 Dr_Drache joined #fuel
13:19 akasatkin >>> We basiclly seem to have a bridge on every eth in the machine even though one (eth1) is not assigned.  I was wondering how this is done?
13:19 akasatkin "how": what do you mean ? They're created with OVS.
13:20 baboune how do the bridge get created in the Ubuntu install of Mirantis 4.1?
13:20 baboune even on eth interfaces not associated with openstack?
13:20 baboune basically that one was supposed to be used to connect to our "extenal network"
13:21 akasatkin Yes, bridges are created for all NICs now.
13:22 akasatkin You can create a port on that bridge, and assign IP to that port.
13:22 baboune humm.. so how would I setup that bridge for that interface so that it can use a static ip?
13:23 baboune so another bridge on top of the bridge?
13:23 baboune sorry, a port on that bridge in OVS?
13:23 akasatkin yes? OVS port on that bridge
13:24 Ch00k joined #fuel
13:24 baboune can I then use this bridge/port/ip as a novnc_proxy_base_url to redirect the vnc  console?
13:25 baboune I dont see why this bridge is useful on that card?
13:34 baboune ovs-vsctl del-br removes it and let s us use the eth normally...  Can this auto gridge feature be turned offf in Fuel?
13:40 akasatkin yes, you can remove bridge manually. It cannot be turned off in Fuel UI but you can edit "transformations" section of network parameters to remove bridge instantiation using CLI.
13:44 Tonyco joined #fuel
13:49 jobewan joined #fuel
13:59 akasatkin see http://docs.mirantis.com/fuel/fuel-4.1/user-guide.html#understanding-environment-deployment-with-fuel-cli , commands deployment download/upload
14:22 xenolog joined #fuel
14:27 dhblaz joined #fuel
14:44 Ch00k joined #fuel
15:00 IlyaE joined #fuel
15:57 rmoe joined #fuel
15:58 dburmistrov joined #fuel
16:26 vkozhukalov_ joined #fuel
16:27 xarses joined #fuel
16:39 angdraug joined #fuel
17:04 richardkiene_ joined #fuel
17:11 Ch00k joined #fuel
17:17 fandi joined #fuel
17:20 TVR___ quiet today
17:23 Dr_Drache yea
17:23 Dr_Drache i'm still yesting on 4.0
17:25 TVR___ I just got a 10G switch plugged in... will see if I can get better than 125MB /s on my ceph clusters now
17:26 TVR___ *gee, I wonder where my bottleneck was*
17:28 obcecado is it usual to deploy ceph on two nodes? any recommendations? just asking this because of swift's three node requirement
17:29 TVR___ two node ceph will not work as it will not be able to create a quorum for the cluster... 1 or 3+ is needed.
17:29 TVR___ 3 or 3+... not just 3+
17:30 Dr_Drache TVR___, you confused me, and I know what you are saying.
17:30 TVR___ I could confuse Confucius ....
17:30 Dr_Drache obcecado, in ceph, you want an odd # of ceph nodes, 3 or more is reccomended.
17:31 obcecado ok
17:32 obcecado thank you for your input
17:36 MiroslavAnashkin Please do not forget to set proper (3 or higher) Ceph replication factor for your environment on Settings tab
17:38 obcecado the thing is
17:38 obcecado all my storage runs on raid
17:38 obcecado having three nodes running storage nodes
17:39 obcecado will dup my data even more
17:39 Dr_Drache that's one of the draws of ceph.
17:39 obcecado from what i've understood
17:39 MiroslavAnashkin Ceph is raid itself
17:39 obcecado it's what you get in all storage services?
17:40 obcecado swift would do the same, or am i wrong?
17:40 obcecado in terms of data dup
17:40 Dr_Drache swift would still need a backend.
17:40 Dr_Drache well.... that's not entirely correct.
17:40 dhblaz joined #fuel
17:40 Dr_Drache but yes.
17:40 MiroslavAnashkin Yes, swift is also rais and for both Swift and Ceph hardware RAID is not necessary
17:40 Dr_Drache ceph doesn't need you to raid your disks pre-osd
17:40 MiroslavAnashkin rais/raid
17:41 obcecado yes, but unfortunately it's what i have available
17:41 obcecado to demo/test/tinker :-)
17:41 Dr_Drache MiroslavAnashkin can correct me, but I think it's counter productive to have hardware raid with ceph.
17:41 obcecado ok so, let me put it this way
17:41 obcecado if you have raid, what would be the optimal backend to use?
17:41 obcecado none ? :-)
17:41 MiroslavAnashkin Dr_Drache: Depends on RAID level. Some levels increase read speed
17:42 Dr_Drache MiroslavAnashkin, true.
17:42 TVR___ optimal backend if everything is raid already and you feel good would be a simple LVM raid 0 stripe accross all the volumes
17:43 Dr_Drache raid-0 stripe is ok. but, I forget the speed increase of raid. ceph is 80%.
17:43 obcecado to be honest, i don't think i trust having a streched lvm volume between the hosts running on multiple disk arrays
17:43 TVR___ here is the issue with raid and ceph, swift, of gluster on top.... raid, hardware raid that is, inherently has a penalty for disk integrity.... so it is like using disks, but asking them to run at 80% speed....
17:44 obcecado i understand
17:44 TVR___ ceph, swift, or gluster deals with the replication sets (the reason for raid) but without all the penalties associated with it, and there are gains from spindle count and scaling horizontally ..
17:45 obcecado doing the same tasks at different layers
17:45 obcecado btw, anyone here attended mirantis bootcamp @ paris last month?
17:45 TVR___ raid replicates parity while clusters replicate data.. (unless we are talking raid 1)
17:46 Ch00k joined #fuel
17:46 TVR___ raid does not scale past the chassis while clustering scales horizontally... it is just a larger scale way to preserve data
17:56 IlyaE joined #fuel
18:06 xarses TVR___: obcecado: Dr_Drache: the number of OSD nodes required at deployment is the same (or larger than) the replica count you want. You where discussing quorum, which is only relevant to the monitors. For monitors, fuel deploys them on each controller, if you want redunancy, just as with the controllers, you need at least 3 and should keep an odd number to prevent split brains
18:08 Dr_Drache xarses, so true.
18:11 e0ne joined #fuel
18:17 xarses obcecado: Dr_Drache, correct, you do not want to use raid with ceph, the whole point is to put one OSD per device, ceph will handle "raid". There are some conditions, like when you have thousands of OSD's that reducing the number of OSD's will improve placement calculation. In which case, you would create raid 0's of disks to reduce OSD count while not duplicating data protection. This does however increase the recovery time and impact mtbf calculations
18:18 xarses similar points can be made with swift as well since it handles object replication
18:25 vk joined #fuel
18:25 TVR___ so if I misconfigure my VLANS so management and storage cannot talk to each other, why won't fuel install my cluster?
18:25 TVR___ heh
18:26 TVR___ after my install failed I saw the VLANs for my 10G switch were wrong...
18:26 TVR___ heh
18:26 TVR___ try again..
18:35 dhblaz xarses: Do you have any suggestions regarding the neutron router issue I discovered yesterday?
18:49 MiroslavAnashkin TVR___: Network check may take hours or days in case you configured many VLAND and use hundreds of nodes. It is the reason we didn't made it mandatory.
18:49 MiroslavAnashkin VLANs
19:05 rvyalov joined #fuel
19:09 designated joined #fuel
19:18 TVR___ cool.. good to know
19:19 TVR___ so, I want to expand my ceph cluster with disks on the nodes already in the cluster...
19:19 TVR___ I basically want to do a ceph-deploy osd create on them...
19:19 TVR___ doesn't seem to work...
19:20 TVR___ ceph-disk: Error: Device /dev/sdf3 is in use by a device-mapper mapping (dm-crypt?): md0
19:20 TVR___ thoughts?
19:22 Ch00k joined #fuel
19:22 designated finally got a successful deployment using fuel 4.1 on ubuntu12.04, HA with neutron/vlans but the "Check network connectivity from instance via floating IP" test is failing and I'm showing Admin state up but Status down for all of the network ports on my External network.  Any ideas why or where to begin looking?
19:24 designated in horizon under my external network it shows "Network Type: flat"  that's obviously wrong
19:35 dhblaz designated: I have found that the "Status down" for float ports is normal.
19:36 dhblaz If you look carefully at the test you can see what step it failed on along with a reason and that might shed some light on the problem
19:36 dhblaz How many floats do you have?
19:37 designated it's failing at check that public ip 8.8.8.8 can be pinged from instance.
19:37 designated 1 float
19:37 designated hosts aren't getting dhcp and not able to talk to assigned floating ip
19:38 designated there is definitely a network issue somewhere, never seen hosts not be able to pull dhcp
19:39 e0ne joined #fuel
19:44 e0ne_ joined #fuel
19:56 rvyalov joined #fuel
20:20 rvyalov joined #fuel
20:25 designated When the external network got created it shows: Network Type: flat
20:26 designated now i can't delete it to make a new one of type: vlan
20:27 justif ubuntu deploy with 1 controller, 3 ceph, 2 compute - Compute nodes and controller node shows the linked error - http://i.imgur.com/Z62M5BV.png
20:28 Dr_Drache damn
20:28 Dr_Drache I can't even deploy ubuntu on 4.1, wish I could help debug
20:29 justif cant type S or M
20:32 justif Also side question not related to my issue. Is there a way to integrate an iSCSI SAN such as a Equallogic shelf as the backend storage?
20:34 bookwar left #fuel
20:35 designated i definitely think there is an issue when 4.1 deploys that external network.  It should be type: vlan not flat right?
20:38 dhblaz One float isn't enough
20:41 dhblaz Also this will be a problem for you if you have very limited number of floats:
20:41 dhblaz https://bugs.launchpad.net/bugs/1288547
20:41 dhblaz designated: I think that the neutron router uses a float for it's IP address
20:41 dhblaz So you wouldn't have any left for your instances
20:42 dhblaz If this is an HA install check crm status for problem from one of the controllers
20:44 dhblaz I don't have a 4.1 install but in 4.0 sometimes crm resource restart p_neutron-dhcp-agent can help
20:46 dhblaz If you want to see if the dhcp requests are getting to your head node you would do it like this:
20:46 designated i have more than one floating ip, i thought you were asking about float pools
20:46 designated i have allocated over 100 floating ip addresses
20:53 designated after deployment i see the following: neutron.common.legacy [-] Skipping unknown group key: firewall_driver
20:53 designated followed by:
20:53 designated neutron.api.rpc.agentnotifiers.dhcp_rpc_agent_api [-] No DHCP agents are associated with network '13c2e11e-2890-4d06-99e4-64f8ee1928c1'. Unable to send notification for 'network_create_end' with payload: {'network': {'status': 'ACTIVE', 'subnets': [], 'name': u'net04_ext', 'provider:physical_network': u'physnet1', 'admin_state_up': True, 'tenant_id': u'93fc77ff09e6486cbb1b3b172bb6dcef', 'provider:network_type': u'flat', 'router:exte
20:58 dhblaz what do you see if you do (from a controller) source openrc; neutron agent-list
20:58 dhblaz Do you see a DHCP agent listed?
20:58 dhblaz It looks like this is a flat network like everything else is reporting. :(
20:59 designated I have openvswitch agents listed, a L3 agent, and a DHCP agent all show admin_state_up TRue
21:00 designated i think the external network just got created incorrectly as a flat network instead of a vlan network
21:02 designated a neutron net-show id gives me: | provider:network_type     | flat
21:08 designated have any of the devs been around today?
21:14 justif both xarses and Miroslav were around earlier
21:15 * xarses hides
21:15 justif lol
21:15 * xarses has no clue whats going on
21:17 rmoe is there a segmentation ID for that network?
21:18 designated rmoe: are you asking about vlan ids?
21:18 rmoe yes
21:18 designated no because it's defined as type: flat
21:19 designated all over networks i create come up as type: vlan
21:19 designated over == other
21:20 designated do you know if there is a way to change it to vlan?
21:22 rmoe you can delete and recreate the network
21:27 designated I've yet to figure out how to delete it
21:27 rmoe in Fuel did you set a VLAN on your public network on the settings page?
21:27 designated it tells me there are ports connected, then i try to delete ports
21:28 designated no
21:28 rmoe but your network requires all external traffic to be tagged?
21:28 obcecado thank you for your reply xarses
21:28 designated rmoe no
21:28 xarses obcecado: yw
21:29 designated rmoe: i got this error following deployment. neutron.common.legacy [-] Skipping unknown group key: firewall_driver
21:29 rmoe ok, so your problem is that your hosts aren't getting DHCP addresses? (I'm reading the scrollback, just want to make sure I know what's going on)
21:29 designated rmoe correct
21:29 rmoe ok
21:29 rmoe which subent are you attaching your VMs to?
21:29 rmoe subnet
21:30 designated i created a new network and connect a router between my new network and the external network
21:30 designated neutron.api.rpc.agentnotifiers.dhcp_rpc_agent_api [-] No DHCP agents are associated with network '13c2e11e-2890-4d06-99e4-64f8ee1928c1'. Unable to send notification for 'network_create_end' with payload: {'network': {'status': 'ACTIVE', 'subnets': [], 'name': u'net04_ext', 'provider:physical_network': u'physnet1', 'admin_state_up': True, 'tenant_id': u'93fc77ff09e6486cbb1b3b172bb6dcef', 'provider:network_type': u'flat', 'router:exte
21:30 rmoe and you enabled DHCP on the subnet associated with your new network?
21:33 designated correct
21:35 rmoe I thnk that message is fine, it looks like it's saying that no DHCP agents are associated with net04_ext which is correct because that network doesn't enable dhcp
21:35 designated but i get the same message when i create a new network
21:36 designated neutron.api.rpc.agentnotifiers.dhcp_rpc_agent_api [-] No DHCP agents are associated with network 'a0387cd3-c7c2-428f-8804-e2e7334a1c37'. Unable to send notification for 'subnet_create_end' with payload:
21:36 rmoe ok
21:36 rmoe let me check into this on one of my deployments
21:36 designated i do a neutron agent-list and show my DHCP agent
21:37 designated then i do a neutron net-list-on-dhcp-agent id
21:37 designated a0387cd3-c7c2-428f-8804-e2e7334a1c37 | net-1 | ed7b3732-1a44-4870-8736-9784299c96fd 192.168.200.0/24
21:37 vkozhukalov_ joined #fuel
21:43 rmoe those messages are fine, they're just about notifications http://docs.openstack.org/admin-guide-cloud/content/section_adv_notification.html
21:43 rmoe I see the same messages on my deployments but DHCP works for VMs on any new networks I create
21:43 rmoe are there any other errors in the dhcp agent log file?
22:08 designated not that i see
22:11 designated if i deploy an instance and do an ifconfig, i don't see an ip address listed and cannot ping the default gateway
22:22 warpig joined #fuel
22:53 justif joined #fuel
22:54 justif lol, all nodes at ready but deployment failed due to timeout and cant deploy again and no errors in the logs
22:56 justif can log into horizon and default testvm seems to be stuck "Saving"
22:58 justif cant create volume so ceph might be borked
23:21 e0ne joined #fuel
23:28 xarses justif: ceph -s from one of the controllers
23:29 justif alaready reset the environment but it showed healthy_err irrc and no OSDs
23:29 xarses yep, that would cause the issue
23:29 justif waiting for 2 more nodes to come back after resetting and going to try again
23:30 xarses sounds like a networking issue on the management network between the ceph mons (controllers) and the nodes with the ceph osd role
23:30 e0ne joined #fuel
23:30 justif 5 total machines 3 ceph+controller and 2 compute
23:30 xarses you should check the storage network too
23:31 justif with neutron gre with teh ovs hard splinters
23:31 xarses centos?
23:31 justif yea
23:31 xarses ubuntu still flakey for you?
23:32 justif yea still fails after the fsck looking for the glance volume and nova volumes
23:32 justif same ss from last night
23:32 justif went as far as nuking all the partitions on all the nodes
23:33 justif the fedora kernel does not have the bnx2 firmware for my nics
23:33 justif else i would use that
23:35 e0ne_ joined #fuel
23:39 justif brb grabbing foods
23:42 richardkiene joined #fuel

| Channels | #fuel index | Today | | Search | Google Search | Plain-Text | summary