Perl 6 - the future is here, just unevenly distributed

IRC log for #fuel, 2014-01-21

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

All times shown according to UTC.

Time Nick Message
23:00 e0ne joined #fuel
23:09 vk joined #fuel
23:14 xarses its supposed to prevent it from loosing track of the PID and killing it thinking it's down
23:41 jaypipes joined #fuel
23:55 e0ne joined #fuel
23:57 e0ne joined #fuel
23:58 e0ne joined #fuel
00:00 e0ne joined #fuel
01:25 richardkiene_ is the br-ex interface always configured using the Public interface?
01:25 richardkiene_ I'm not seeing an easy way to verify in /etc/network/interfaces.d/
01:57 e0ne joined #fuel
02:35 vkozhukalov joined #fuel
02:48 xarses joined #fuel
02:57 e0ne joined #fuel
02:59 rmoe joined #fuel
03:02 IlyaE joined #fuel
03:42 richardkiene joined #fuel
03:57 e0ne joined #fuel
04:12 besem9krispy joined #fuel
04:24 jouston joined #fuel
04:34 dhblaz joined #fuel
04:36 rmoe joined #fuel
04:39 ArminderS joined #fuel
04:57 e0ne joined #fuel
05:19 jhurlbert_ joined #fuel
05:26 IlyaE joined #fuel
05:43 ArminderS joined #fuel
05:57 e0ne joined #fuel
06:26 rongze joined #fuel
06:43 ArminderS joined #fuel
06:57 e0ne joined #fuel
07:04 e0ne joined #fuel
07:30 IlyaE joined #fuel
07:49 mihgen joined #fuel
08:13 mrasskazov joined #fuel
08:20 mihgen_ joined #fuel
08:29 vkozhukalov joined #fuel
08:35 mrasskazov joined #fuel
09:02 miguitas joined #fuel
09:12 mrasskazov joined #fuel
09:19 teran joined #fuel
09:29 mihgen joined #fuel
09:53 e0ne joined #fuel
10:04 teran joined #fuel
10:09 getup- joined #fuel
10:12 besem9krispy joined #fuel
10:33 anotchenko joined #fuel
10:36 mrasskazov joined #fuel
10:37 vk joined #fuel
10:54 vk joined #fuel
11:15 e0ne joined #fuel
11:22 rongze joined #fuel
11:35 rongze joined #fuel
11:48 anotchenko joined #fuel
12:09 mattymo joined #fuel
12:35 rongze joined #fuel
12:37 anotchenko joined #fuel
13:03 mihgen joined #fuel
13:21 vk joined #fuel
13:27 e0ne_ joined #fuel
13:28 teran_ joined #fuel
13:35 rongze joined #fuel
13:43 vk joined #fuel
14:24 e0ne joined #fuel
14:35 rongze joined #fuel
14:41 xdeller joined #fuel
14:52 Arminder- joined #fuel
14:57 ArminderS joined #fuel
14:57 IlyaE joined #fuel
15:12 rongze joined #fuel
15:23 mihgen joined #fuel
15:42 richardkiene_ joined #fuel
15:49 richardkiene joined #fuel
16:06 rmoe joined #fuel
16:19 syt Has anyone done a default fuel rollout using GRE networking and CEPH as storage backend where all the openstack storage functionality actually works?
16:19 syt We're havind some troubles with resizing, accessing the object store, creating volumes, etc etc..
16:19 syt sometimes it works, sometimes it doesn't
16:40 teran joined #fuel
16:57 rongze joined #fuel
17:07 vkozhukalov joined #fuel
17:12 ArminderS in generate_network_config.rb, with nic-bonding, where there is check to see interfaces parameter, why do we have condition to fail if interfaces size !=2 instead of >=2 ?
17:12 ArminderS the check fails if we have more than 2 NICs added up in a bond
17:12 ArminderS shouldn't that be valid thing to do?
17:15 MiroslavAnashkin ArminderS: It is due to currently not fixed OpenVSwitch issue. OVS works unstable if bond includes more than 2 interfaces.
17:16 ArminderS so which means its going to be a bottleneck, any work arounds?
17:18 MiroslavAnashkin ArminderS: Workaround is to fix OVS first.
17:19 ArminderS so thats why its hard-coded to a value of 2
17:20 ArminderS thanks MiroslavAnashkin
17:20 MiroslavAnashkin ArminderS: Yes, using Linux bonds does not help here as well
17:20 ArminderS also can you confirm that in ceph as backend, the volume traffic travels over management network and ceph osd replication travels over storage network, right?
17:21 MiroslavAnashkin ArminderS: Yes, it should work this way.
17:22 ArminderS thanks for confirming
17:23 ArminderS i used to wonder why nic-bonding used to fail since i was trying to add >2 NICs in a single bond, until i examined the code
17:24 MiroslavAnashkin syt: such configuration is being deployed and tested on every internal Fuel iso build.
17:25 angdraug joined #fuel
17:28 rongze joined #fuel
17:28 richardkiene syt: We've deployed that exact setup at my company a few times now. There were some rough patches with the HA install and choosing between CentOS kernel limitations versus Ubuntu configuration bugs, but those were performance limitations that we fixed, and Ceph has always been functional
17:28 xarses joined #fuel
17:29 ArminderS ceph worked good for me too as backends for volumes/images/objects
17:30 syt MiroslavAnashkin: I thought so, we're seeing a lot : nova-glanceclient.common.http ERROR: Request returned failure status. and no clue to what is might be.
17:31 syt richardkiene: it all seems to work, only not all the time.
17:31 richardkiene syt: What type of deployment are you using? HA Controllers? Host OS?
17:31 syt If I create 50 instances with volumes, mayby 10 are created correctly, the others fail.
17:31 syt richardkiene: the 'not-HA' setup using centos
17:33 rmoe joined #fuel
17:34 richardkiene syt: Interesting, we had more than a few bumps in the road with centos (and really only tried it out, we're far more comfortable with Ubuntu)
17:35 rongze joined #fuel
17:44 ArminderS MiroslavAnashkin: I tried to search for that OVS issue but can't find it. Can you point me to that so that I can keep an eye on it?
17:46 ArminderS the ovs man page says it needs atleast 2 interfaces
17:47 syt richardkiene: we're a Ubuntu company, but for some reason ubuntu handles the interfaces differently and the auto provisioning fails, so in order to test Fuel we descided to use CentOS
17:53 syt We're now at the point to do a bare installation by ourselves so we have a better idea of how/where to debug.
17:58 richardkiene syt: There is definitely merit in that
17:58 richardkiene syt: we also went through the pain of a manual installation, and we learned a lot from it
17:58 richardkiene syt: In the long run though, Fuel is insanely helpful
17:59 richardkiene syt: I will also say that if you post what your problems are in here, I've had nothing but great help from the Mirantis crew and everyone else in this room
18:01 teran joined #fuel
18:03 ArminderS now with 2 interfaces, getting this error -> (/Stage[netconfig]/Advanced_node_netconfig/L23network::L2::Bond[bond0]/L2_ovs_bond[bond0]/ensure) change from absent to present failed: Could not set 'present on ensure: can't convert String into Array at /etc/puppet/modules/l23network/manifests/l2/bond.pp:59
18:04 ArminderS i wonder is it due to properties value for add-bond?
18:15 MiroslavAnashkin ArminderS: Hmm, it more and more looks like a bug in generate_network_config.rb. OVS works incorrect with Linux bonds only and with specific traffir only. So, you may change interfaces size to >=2. There in no such limit for OVS bonds.
18:17 ArminderS so by editing the check from !=2 to >=2, that will work, right?
18:19 MiroslavAnashkin ArminderS: Yes, should work with OVS bonds definitely. May work with Linux bonds in most cases.
18:24 ArminderS great
18:24 ArminderS since fuel uses OVS bonds, so should work
18:33 MiroslavAnashkin syt: What Tunnel ID range have you set for OpenStack environment, where you experiencing failures with 50 instances creation?
18:34 MiroslavAnashkin syt: You may also share diagnostic snapshot, so we could check your network settings.
19:01 IlyaE joined #fuel
19:10 syt richardkiene: MiroslavAnashkin: Unforunately I'll have to get back to you on Thursday :S
19:11 e0ne joined #fuel
19:32 IlyaE joined #fuel
20:37 rongze joined #fuel
20:43 e0ne joined #fuel
21:00 vt102 joined #fuel
21:05 angdraug joined #fuel
21:06 e0ne joined #fuel
21:31 e0ne joined #fuel

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