Perl 6 - the future is here, just unevenly distributed

IRC log for #fuel, 2015-01-27

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

All times shown according to UTC.

Time Nick Message
00:34 adanin joined #fuel
00:41 moizarif cc;"????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????????
01:05 jpf joined #fuel
01:35 LiJiansheng joined #fuel
01:52 MarkDude joined #fuel
02:01 rmoe joined #fuel
02:16 mattgriffin joined #fuel
02:24 xarses joined #fuel
02:54 fandi joined #fuel
02:56 fandi joined #fuel
02:59 fandi joined #fuel
03:07 fandi joined #fuel
03:11 fandi joined #fuel
04:12 mattgriffin joined #fuel
04:34 moizarif joined #fuel
04:39 teran joined #fuel
05:40 teran joined #fuel
05:58 Arminder joined #fuel
05:59 Arminder- joined #fuel
06:03 Arminder joined #fuel
06:04 thumpba joined #fuel
06:05 ebogdanov joined #fuel
06:05 Arminder- joined #fuel
06:08 Arminder- joined #fuel
06:08 Miouge joined #fuel
06:19 teran joined #fuel
06:32 Arminder joined #fuel
06:48 sambork joined #fuel
07:11 avgoor joined #fuel
07:17 ebogdanov joined #fuel
07:22 aliemieshko_ joined #fuel
07:35 Miouge joined #fuel
07:36 aliemieshko_ joined #fuel
07:37 strictlyb joined #fuel
07:52 adanin joined #fuel
07:54 corepb_ joined #fuel
07:55 e0ne joined #fuel
07:57 saibarspeis joined #fuel
08:09 azemlyanov joined #fuel
08:11 dklepikov joined #fuel
08:47 sambork joined #fuel
08:50 baloney_ joined #fuel
08:58 mescanef joined #fuel
09:21 ebogdanov joined #fuel
09:23 teran joined #fuel
09:25 HeOS joined #fuel
09:26 teran_ joined #fuel
09:27 fandi joined #fuel
09:36 stamak joined #fuel
09:44 baloney_ joined #fuel
09:46 warpc__ joined #fuel
09:53 sambork joined #fuel
09:55 alecv joined #fuel
10:23 baloney_ joined #fuel
10:33 dancn joined #fuel
10:47 ChrisNBlum joined #fuel
10:47 e0ne joined #fuel
10:57 ddmitriev joined #fuel
10:59 Miouge championofcyrodi: Ok pretty cool, so it’s linux based VDI/TS ?
11:19 monester_laptop joined #fuel
11:20 sambork joined #fuel
11:27 sambork1 joined #fuel
12:21 baloney_ joined #fuel
12:29 subscope joined #fuel
12:39 baloney_ hi
12:41 baloney_ I have set up a stack with Neutron/GRE, 2 NICs. Eth0 is for admin/PXE and eth1 for everything else. Now, the fuel web admin is available through eth1, but the openstack admin only listens to the internal network and the admin one. Do I need to manually setup the routing or did I misconfigure during the install?
12:41 baloney_ eth0 is separated from the world atm
12:51 subscope joined #fuel
13:19 hyperbaba joined #fuel
13:22 hyperbaba Hi there, one of my osd disk somehow got 100% full. Osd won't start on it. Error is genericfilestorebackend(/var/lib/ceph/osd/ceph-0) detect_features: failed to write to /var/lib/ceph/osd/ceph-0/fiemap_test: (1) Operation not permitted
13:23 hyperbaba Can anyone help me in resolving this problem. On the other hand entire cluster is stuck with some 180 pgs stuck / unclean
13:26 Longgeek joined #fuel
13:29 NuClEaR__ joined #fuel
13:38 teran joined #fuel
13:41 championofcyrodi Miouge: correct.  Using LTSP I boot Intel NUCs (w/ core i3 and 16GB RAM) as FAT clients.  The excess RAM allowes the root filesystem to be cached in read-only mode.  then using UnionFS, a writeable tmpfs is layered over top.  This creates a usable linux distro that is reset on reboot when the tmpfs is reset.  I then use the clients to NX to the headless openstack VMs.
13:41 championofcyrodi with LDAP authentication everywhere
13:42 championofcyrodi also, on the fat client, the user's home folder is network mounted at login
13:49 MiroslavAnashkin hyperbaba: Please set lower weight to the 100% full OSD first. Much lower.
13:51 tzn joined #fuel
13:52 MiroslavAnashkin Then, if OSD start - wait until Ceph finish data redistribution,
13:54 emagana joined #fuel
13:59 hyperbaba MiroslavAnashkin: weight of that disk is 0.23 ... the other are 0.91
13:59 MiroslavAnashkin baloney_: By default Fuel UI is accessible on all master node NICs. Eth1 for everything else means eth1 will be used for everything on OpenStack nodes, but will not be used on master node at all or you may get access to Fuel UI through its network.
13:59 MiroslavAnashkin hyperbaba: Please set weight to 0.01
14:00 hyperbaba MiroslavAnashkin: ok. But it still does not start there is no space on device for the initial test
14:07 MiroslavAnashkin Then you may delete some data on mounted full OSD manually - ceph should have a copy of deleted placement groups on another OSDs. Then start this OSD and set lower weight.
14:08 baloney_ MiroslavAnashkin: Fuel UI is ok, I was talking about OpenStack UI. I am currently trying to understand GRE with respect to VLAN because compute nodes can't reach the controller..
14:08 MiroslavAnashkin BTW, how many OSDs do you use?
14:09 rbowen joined #fuel
14:10 MiroslavAnashkin baloney_: It depends on which NIC you have configured for private OpenStack network before cluster deployment.
14:23 baloney_ MiroslavAnashkin: I set up eth0 (admin/PXE) between all the machines, and eth1 for whatever else. Installation worked like a charm. Assigning to nodes and everything. So, the private network would be eth0 I suppose.
14:32 MiroslavAnashkin No. Private network is not accessible at all - it is virtual network existing between OpenStack tenants inside Neutron only. So, it is Admin/PXE network is running on eth0. Nodes use it for NTP and some external routing.
14:40 mattgriffin joined #fuel
14:41 Miouge joined #fuel
14:54 baloney_ MiroslavAnashkin: so the private network is eth0 like I said? cause you started the reply with "No" and then you said "it is Admin/PXE running on eth0".
14:55 baloney_ MiroslavAnashkin: not accessible could mean br-mgmt and friends
14:59 monester_laptop joined #fuel
15:02 julien_ZTE joined #fuel
15:03 tzn joined #fuel
15:03 teran_ joined #fuel
15:05 teran_ joined #fuel
15:07 jobewan joined #fuel
15:08 teran joined #fuel
15:13 emagana joined #fuel
15:16 julien_ZTE joined #fuel
15:18 mpetason joined #fuel
15:19 championofcyrodi any tips/notes floating around on decommissioning a single controller in an HA cluster w/ 3 controllers?
15:25 blahRus joined #fuel
15:29 championofcyrodi MiroslavAnashkin: When you say, "create and export new image from an instance, then import the image on new environment."  would this be done using cinder's backup-create and backup-restore options?
15:46 adanin joined #fuel
15:49 mescanef joined #fuel
15:56 omolchanov joined #fuel
15:57 jaypipes joined #fuel
16:04 NoisyNoise joined #fuel
16:05 fandi joined #fuel
16:07 dklepikov joined #fuel
16:09 fandi joined #fuel
16:11 mattgrif_ joined #fuel
16:12 NoisyNoise hello all
16:13 tobiash_ joined #fuel
16:13 seeg joined #fuel
16:14 NoisyNoise I have a quick question about HA in Fuel 6.0 : 3 controllers are mandatory for HA, is there a way to bypass this limit and define 2 controllers as requirement ?
16:14 NoisyNoise and I will use a galera arbitrator as third player for the the quorum
16:14 ChrisNBlum joined #fuel
16:17 evg NoisyNoise: hello, you can deploy the only one controller in HA mode and add others later
16:17 e0ne joined #fuel
16:17 meow-nofer__ joined #fuel
16:19 NoisyNoise evg : ok thanks. so If I install only 2 controllers in HA and I deploy one galera arbitrator by my hands, Fuel won't complain about the infrastructure and I could use it without problem ?
16:23 adanin_ joined #fuel
16:39 emagana joined #fuel
16:43 MiroslavAnashkin baloney_: There is OpenStack Private network in Nova Network and Neutron - it exists not in all configurations but there is. And there is Admin/PXE network. I assume you call this Fuel Admin/PXE network as private, which is incorrect.
16:46 monester_laptop joined #fuel
16:46 MiroslavAnashkin championofcyrodi: Yes, with nova image-create
16:47 MiroslavAnashkin championofcyrodi: http://docs.openstack.org/user-guide/content/cli_migrate_instances.html
16:47 championofcyrodi MiroslavAnashkin: thanks, was trying: http://docs.openstack.org/admin-guide-cloud/content/volume-backup-restore-export-import.html
16:49 MiroslavAnashkin About controller decommissioning in Fuel - simply mark the controller in Fuel UI, click Delete (sets node to pending deletion), then click Deploy Changes (it updates configs on remained controllers and removed pending nodes)
16:51 championofcyrodi MiroslavAnashkin: ty again.
16:52 MiroslavAnashkin baloney_: As of computes unable to access controllers - please check on the compute node, if GRE tunnels created with `ovs-vsctl show`
16:52 mattgriffin joined #fuel
16:54 emagana joined #fuel
16:56 emagana_ joined #fuel
16:56 e0ne joined #fuel
16:57 clauded joined #fuel
17:00 clauded Hi. Running Fuel 5.1 with nova vlan configuration and multiple compute hosts. If VMs are on different compute hosts, I can't ping them with their dns name (only with their ip address). Is this considered normal with fuel. I tought it was taken care with this fix in openstack: https://review.openstack.org/#/c/16153/
17:02 e0ne_ joined #fuel
17:03 Miouge joined #fuel
17:04 julien_ZTE joined #fuel
17:05 t_dmitry joined #fuel
17:16 MiroslavAnashkin clauded: If you use Neutron+VLAN - there is only blueprint to add DNS name resolution to Neutron in future versions. if you use Nova Network - DNSMasq is working as DNS server, but needs additional configuration, like external hosts file.
17:17 clauded MiroslavAnashkin: I'm using Nova network
17:29 clauded MiroslavAnashkin: I can see the file addn-hosts (/var/lib/nova/networks/nova-br1000.hosts) but how is this file managed? Manually?
17:30 jaypipes joined #fuel
17:32 MiroslavAnashkin clauded: Probably manually or by some script, taking instance names from DB and populating the given hosts file...
17:41 rmoe joined #fuel
17:49 thumpba joined #fuel
17:58 angdraug joined #fuel
17:58 ebogdanov joined #fuel
18:11 xarses joined #fuel
18:34 clauded joined #fuel
18:35 mattgriffin joined #fuel
18:38 e0ne joined #fuel
18:49 kupo24z joined #fuel
18:49 Miouge joined #fuel
19:09 championofcyrodi can fuel support both ceph and local disk (hypervisor disk) hosted volumes at the same time?
19:09 championofcyrodi so we can use ceph for snapshots/images, but raw disk for running instances?
19:11 Miouge championofcyrodi: you have different deployment options for each storage
19:12 Miouge glance backend, can be swift or ceph. cinder backend can be either LVM or ceph, nova ephemeral storage can be local disk or ceph
19:14 championofcyrodi hmm... well our ceph is slow... i'm thinking about local disk for ephemeral storage and ceph for everything else.
19:14 Miouge Specificly “Ceph RBD for ephemeral volumes (Nova)”
19:15 Miouge That is the setup I am testing (cinder with ceph, but local disk ephemeral)
19:16 Miouge The “problem” I have with that is that local disk ephemeral is recomended to use qcow2 images, but ceph is recommended to use raw images
19:17 championofcyrodi we are also concerned with being able to live migrate though...
19:17 teran joined #fuel
19:18 championofcyrodi our understanding is that you cannot live migrate an ephemeral instances?
19:19 championofcyrodi for us, a 'master' node in a cluster is the most complex to configure.  so we would like to be able to perform snapshots, migrates etc.  so we are thinking it may be acceptable to run a master node on ceph, then all the workers as ephemeral instances.
19:20 Miouge I have not tested it (with fuel) but you should be able to live migrate with ephemeral storage
19:21 Miouge The downside is that it will migrate the whole disk also (similar to vMotion storage) so its takes longer
19:22 Miouge Makes sense to use cinder (ie ceph) if you need persistence
19:25 daniel3 joined #fuel
19:27 angdraug joined #fuel
19:29 fandi joined #fuel
19:33 championofcyrodi well... i guess were just trying to figure out the most effective way to utilize the 6x256GB SATA II SSDs. spread across 3 nodes (2 per node)
19:33 championofcyrodi problem with using them for ceph is that the network is 1Gbps...
19:33 championofcyrodi so it seems wasteful.
19:37 e0ne joined #fuel
19:37 championofcyrodi problem with using them for ephemeral storage is that if someone needs an instance w/ 200GB, then there goes the ssd storage...
19:38 championofcyrodi i'm thinking we will push to run the VM off SSD, and use CEPH for a data mount point...
19:38 championofcyrodi / - root - 20GB SSD
19:38 championofcyrodi /data - dataforapps - 200GB Ceph
19:45 ilbot3 joined #fuel
19:45 Topic for #fuel is now Fuel 5.1.1 (Icehouse) and Fuel 6.0 (Juno) https://software.mirantis.com | Fuel for Openstack: https://wiki.openstack.org/wiki/Fuel | Paste here http://paste.openstack.org/ | IRC logs http://irclog.perlgeek.de/fuel/
19:46 xarses Miouge: you can not live migrate non-shared disk, you can do an offline migration though
19:48 angdraug joined #fuel
19:52 Miouge xarses: so you would need NFS mount on the ephemeral disk or similar
19:52 Miouge ?
19:52 xarses Miouge: any similar shared storage provider, yes
19:53 xarses nfs, ceph, netapp(nfs), etc...
19:55 teran_ joined #fuel
19:56 Miouge xarses: Ok, thanks!
19:57 stamak joined #fuel
19:58 Miouge championofcyrodi: you control the ephemeral disk size with quota & flavors so you could size it to fit the local SSD
19:58 Miouge Moreover if a user needs more capacity cinder/ceph is still an option
20:00 Miouge What I see people using ephemeral storage for stateless instances (proxy, web server, load balancer, app server, …), and cinder for stateful instances (database, …)
20:03 xarses and / root volumes
20:05 teran joined #fuel
20:05 daniel3 joined #fuel
20:10 ebogdanov joined #fuel
20:14 Miouge xarses: sure
20:22 Samos123_ joined #fuel
20:33 rvyalov joined #fuel
20:33 izinovik joined #fuel
20:33 vkramskikh joined #fuel
20:34 mihgen joined #fuel
20:35 MorAle joined #fuel
20:36 meow-nofer joined #fuel
20:37 evg joined #fuel
20:38 sbog joined #fuel
20:38 MiroslavAnashkin joined #fuel
20:39 akislitsky joined #fuel
20:40 apalkina joined #fuel
20:41 holser joined #fuel
20:43 championofcyrodi xarses: when you say "and" root volumes... do you mean cinder for root or ephem?
20:51 geekinutah joined #fuel
20:52 geekinutah guys, how difficult https://blueprints.launchpad.net/fuel/+spec/slave-nodes-change-pxe-iface would be to implement?
20:52 * geekinutah pokes angdraug
20:53 * xarses pokes geekinutah
20:53 xarses what do you mean, what is your problem?
20:54 proberts joined #fuel
20:55 geekinutah proberts can elaborate
20:55 geekinutah apparently he's personally ran into this a few times
20:55 proberts Customer has standard builds where eth0 is reserved for thei rmanagement traffic
20:56 proberts guys I'm getting kicked out of a conf room
20:56 proberts will respond in a few
20:57 xarses the pxe bootstrap interface will be determined automatically from the interface that registers the node with fuel, if you dont wan't it to use eth0 don't have plumb the port to fuel
20:58 geekinutah so all the customer has to do is change their PXE boot NIC order?
21:01 championofcyrodi our slave seems to just choose the eth adapter that is already connected to the LAN with the dhcpd/tftpd pxe support: http://imgur.com/frA5blv
21:01 championofcyrodi eth6 in that case ^
21:18 HeOS joined #fuel
21:27 xarses joined #fuel
21:27 tzn joined #fuel
21:40 proberts joined #fuel
21:41 proberts xarses: going back to the Fuel PXE NIC (https://blueprints.launchpad.net/fuel/+spec/slave-nodes-change-pxe-iface)
21:41 proberts Customer is using eth0 as a standard management port across their datacenter
21:42 proberts it would be great if Fuel would allow you to move where admin/pxe traffic is passed on the slave nodes (compute, controller, mongo, etC)
21:45 xarses proberts: i'm still confused with the problem here
21:46 xarses proberts: your saying that both eth0 and eth1 receive the pxe network?
21:46 xarses from fuel as their default network
21:46 proberts As cabled, physical NIC2 (or eth1) is to be used for PXE
21:47 proberts in Fuel, that node will successfully bootstrap, but it will hardcode the PXE/Admin interface to eth0 or physical NIC 1
21:47 xarses no, it's not hard-coded to eth0 since ~4.1
21:48 xarses it uses whever interface it can find that matches fuel's network, or the interface with the default route if none can be found
21:48 xarses if eth0 isn't supposed to be used, it should not have the fuel pxe traffic, eth1 should have it and it would naturally switch
21:49 proberts It did not
21:50 daniel3 joined #fuel
21:50 proberts NIC2/eth1 was cabled for PXE only, NIC1/eth0 was cabled for management traffic
21:51 xarses what fuel version is this?
21:53 proberts 6
21:59 proberts joined #fuel
22:01 xarses proberts: rmoe suggested that you compare the mac on eth0 and confirm if its nic0 or nic1. We've checked the code and confirmed that it will pick whichever interface dialed in. The only way that it should be picking eth0 is that eth0 is the pxe network
22:02 xarses I'm going to re-validate that against a node i have running
22:08 daniel3 joined #fuel
22:11 daniel3 joined #fuel
22:21 xarses proberts: yep, confirmed that it will move to whichever eth it pxe boots from
23:00 proberts xarses: were you testing on baremetal?
23:01 proberts I'm going to duplicate it here in my lab to triplecheck
23:26 angdraug joined #fuel
23:27 ebogdanov joined #fuel
23:34 jpf That might be good to address in the Mirantis Fuel docs, since it seems the question has come up a few times now
23:34 xarses I think I will convert it to a lp question/answer

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