=== LoRez_ is now known as l === l is now known as Rez === Rez is now known as LoRez [08:55] Bug #1549174 opened: Spurious failure in TestDNSDynamicIPAddresses.test_bind_configuration_includes_dynamic_ips_of_deployed_nodes [09:07] Bug #1549174 changed: Spurious failure in TestDNSDynamicIPAddresses.test_bind_configuration_includes_dynamic_ips_of_deployed_nodes [09:13] Bug #1549174 opened: Spurious failure in TestDNSDynamicIPAddresses.test_bind_configuration_includes_dynamic_ips_of_deployed_nodes [09:27] i'm pretty sure the openstack installer has just hung [09:33] oh, so doing ^c then looking at commands.log helps [09:52] Bug #1549206 opened: Spurious failure in TestDNSConfigModifications.test_add_node_updates_zone [10:04] Bug #1549208 opened: Spurious failure in TestDNSConfigModifications.test_change_node_hostname_updates_zone [10:04] Bug #1549210 opened: bin/test.region: BlockingIOError: [Errno 11] write could not complete without blocking [11:04] Bug #1549230 opened: Spurious failures in DNS configuration tests [13:08] How long should bootstrapping juju take in the openstack installer because I'm about to hit 30 minutes... [13:17] so syslog is full of error (network unreachable) resolving 'streams.canonical.com/AAAA/IN': 2001:503:ba3e::2:30#53 [13:35] imranh, it should take 5 minutes longer than provisioning the machine in MAAS [13:38] kiko: ok so something is wrong [13:39] imranh, yes, but I'm not sure that error matters unless you are on an IPv6-only network [13:39] imranh, if you want to kill the bootstrap and re-run with a --debug it will give us better insight [13:39] imranh, has maas actually provisioned the bootstrap node? [13:39] kiko: I'm not, i'm on a ipv4 only network here [13:39] ok [13:40] I've been running the OS installer with "DEBUG_JUJU_BOOTSTRAP=1 openstack-install" [13:40] ah [13:40] it did soemthing [13:40] imranh: juju can't access to streams.canonical.com ? MAAS at no point tries to access streams.canonical.com. That's probably juju trying to access their own streams and failing [13:41] multi.py:206] Failed to get ip directly: [Errno -2] Name or service not known [13:41] imranh: yeah that's what it seems to be [13:41] roaksoax, well, but that's an IPv6 query [13:41] ok [13:42] so i disabled ipv6 completely inbetween saying i saw those errors in ssylog and it seems to be working now [13:42] kiko: so my understanding with the squid bug is that if it finds a link local address, it iwll try ipv6 first. If the other endpoint has 10+ IPv6 addresses, then squid will fail as it will attempt 10 retires [13:42] kiko: if the endpoint were to have 9 IPv6 and 1 IPv4, the request would go through just fine [13:43] kiko: maybe that's the issue here [13:43] oh how interesting [13:43] it's actually a bug? [13:44] kiko: the issue with squid, yes. Because, I guess that squid3 should never try to reach the internet with ipv6 if the machine where it is running only has link local [13:44] kiko: and even if it does, and fails, it never tries IPv4 if the other endpoint has 10+ Ipv6 addresses [13:45] multi.py:165] Finished MAAS step, now deploying Landscape. [13:45] woop [13:45] progress [13:45] now we're talking! [13:45] roaksoax, should we ship our squid with ipv6 disabled? [13:46] roaksoax, and release note that if you are on ipv6 enable it? [13:51] kiko: no, we should ship it ipv6 enabled, but we could default to ipv4 first. However, the issue is being fixed on squid so when the fix lands, the work around won't really matter much [13:51] ah [13:52] gotcha [13:52] is it easy to default to ipv4? [13:58] kiko: yes, just a config option, that would need to get tested first though [13:59] gui.py:267] A fatal error has occurred: 'NoneType' object has no attribute 'info_message' [13:59] :( [14:00] task.py:71] ran off end of task list, can't start Bootstrapping Juju [14:01] hmm [14:01] imranh, what version of juju? [14:01] kiko: should be the latest for 14.04 one sec [14:02] juju-core 1.25.3 [14:04] trying it all again... [14:07] imranh, that looks like an honest bug.. in landscape? [14:07] imranh, I'd like to know what happens before that? [14:08] kiko: well it ran the landscape configure command with my details that I had entered, then the openstack ui crapped itself (for lack of better phrasing) and i saw those lines in the commands.log file [14:09] s/openstack ui/the openstack installer ui/ [14:10] I'm trying again now [14:10] if it happens I'll report a bug [14:10] hmm [14:10] But i kinda need to get this setup [14:17] here we go [14:23] Bug #1549303 opened: DeprecationWarning: twisted.internet.task.LoopingCall.deferred [14:29] any maintainer online ? [14:30] yes [14:30] imranh, it worked? what changed? [14:31] kiko: just got to the point where it error'd last time [14:31] SUCESS [14:32] :D [14:32] heh [14:33] Sorry it didn't error again, it'd love to help debug the error i got [14:34] once i have this setup I might try and reproduce it in a VM [14:34] question for you kiko , how do I create a partition in curtin to use all the space without specifying a size of a partition [14:34] imranh, and nothing changed? mm [14:35] mag009, once you commission you do know the size of the drive [14:35] how do I pass it to curtin in a variable? [14:35] mag009, but smoser will know the answer to that question (as might blake_r) [14:35] it's a good question [14:35] kiko: I changed nothing apart from wiping the .cloud-install dir after running openstack-install -u [14:36] from my understanding it's only a template that you hardcode the value which is not dynamic at all. [14:36] unless i'm missing something [14:36] mag009, you're wanting to do that with curtin rather than with maas ? [14:36] smoser, I think so [14:37] well the device vda from server to server change [14:37] I'm trying to setup a generic template lvm for all my baremetal [14:37] sda to be the os in lvm ideally I'd like to be able to use : /dev/sda -> lvm [14:37] this is maas 1.9 ? or 1.8 [14:38] but as I can see you are not doing a pvcreate [14:38] so you expect a partition in order to call vgcreate directly [14:38] It wouldn't be hard tho to add a pvcreate function.. [14:39] it'd like to mention that this is for a custom image :) [14:39] maas does not support custom image yet for partitioning if I'm not mistaken [14:40] so thats why I'm using curtin [14:44] mag009, what do you mean by "custom image for partitioning"? [14:44] smoser, do you get my idea ? [14:45] mag009, i do want to have curtin able to do the sorts of things you're after, but at the moment it isn't really well suited for it. [14:46] what i'd suggest you do is provide curtin config that overrides the 'partitioning' stage, and does whatever you want [14:46] you can then script that with sgdisk or sfdisk or parted or whatever. [14:46] this is also not well documented. :-( [14:46] these are things i'd like to fix, and patches are certainly welcome. [14:46] but its just not there yet. [14:47] if I may propose a simple solution : [14:47] this "/usr/lib/python3/dist-packages/curtin/commands/block_meta.py" [14:47] just before the vgcreate [14:47] I'd like to call a pvcreate [14:48] in this particular case I can use the device in my volume vg [14:57] /in/win 4 [15:24] @smoser this is maas 1.9 ? or 1.8 i'm using maas 1.10 [15:26] :) [15:27] so with the storage config you shoudl be able to acheive what you want [15:27] not really [15:27] I'm looking at it and I must specify a size of my partition [15:27] it is painful probably, but you can go to each node and use that ui to cofigure as you'd like. [15:28] but again I'm using a custom image.. [15:28] which is not supported by the ui [15:28] right ? [15:31] mag009: custom image deployment od not get storage/network config no [15:34] here's my problem that I have : size: 0B if I set the size to zero which is what I want to pass to sgdisk [15:34] 'sgdisk', '--new', '1:2048:2047', '--typecode=1:8300', '/dev/vda' [15:34] I want to have 1:2048:0 [15:35] Bug #1549331 opened: Spurious failure in TestPXEConfigAPI.test_pxeconfig_returns_json [15:38] I'd like to have your input I can submit a patch but if you can tell me if its okay with you [15:39] i'd like to do an exception : if you enter a size of : 0B = use all the device space [15:39] and skip the following : length_sectors + offset_sectors [16:18] smoser, ^^ mag009 [16:19] ? [16:20] i tested my patch and it does work [16:21] roaksoax, custom image deployment really should still get sotrage config [16:22] whether or not it applies it should be beyond maas's care. [16:22] mag009, oh. i see.l [16:23] smoser: is there code in curtin to ensure that the config does not apply ? [16:23] i dont like '0' there being the size. we really should support some way of doing that. [16:23] but i dont knwo that 0 is right. [16:23] how you want to exclude it :) [16:24] 100% ? [16:24] size: 100% [16:24] i dont understand why you chose to use sgdisk vs parted [16:24] cause in the case of parted we could of used the percentage straight into the size value [16:24] smoser: since if we send the storage config, curtin will apply it regardless of the machine. So MAAS doesn't send it to ensure that curtin does things the old way [16:25] i do want to support % at some point in some way. [16:25] roaksoax, yeah. but then it woudl be possible for the image you provide dto have curthooks that would actuallyw ork [16:25] and by not sending it you're explicitly dis-allowing that. [16:27] I just need how you want that.. so my patch get approved.. [16:27] lol [16:27] smoser: right, but if we send the config, curtin will still try to partition. Either way, MAAS does not support doing custom partitioning/networking for custom images, nor other OS's [16:28] but you (or we) shoudl have implemented that better. [16:28] mag009, you can proposed the change to curtin if you want. it wil make us think about it. [16:28] my problem is it's a core feature required by our environment [16:29] and i believe it won't happen quickly if I propose something ;) [16:30] I dont mind coding that part if I know what you guys have in mind for that. [16:38] mag009, I think what smoser is saying is a) do the patch b) put it in the bug and c) we'll look at it for inclusion, possibly proposing a slight change in the config [16:38] mag009, so you should be unblocked for now -- smoser can roll out a new curtin release easily once that's done [16:41] yes Its what I concluded thx [17:15] finally I got a debian install with lvm partition like I wanted [17:15] :D [17:22] mag009, rook [17:35] Bug #1549397 opened: Spurious failure in TestStaticIPAddressManagerMapping.test_get_hostname_ip_mapping_returns_raw_ttl [18:03] is it possible to migrate one maas to another ? [18:03] is there a script to do that? [18:03] maas v 1.9 -> 1.10 [18:07] mag009: you mean upgrade from 1.9 to 1.10 ? you can do that if you wanted to upgrade to xenial [18:07] mag009: 1.10 is just basically maas 1.9, but running with django 1.8 and python3 [18:09] i have 2 maas installation [18:09] I want to import from my other maas instance which is v1.9 [18:09] to my new one 1.10 [18:09] mag009: ah, so we dont have a migration script because we support upgrade path [19:52] using 1.8.0, attempting to update the 'name' of a node-group via REST using the maasclient Python module. the put is returning 200 OK, by the name is not being updated? nothing in the logs, thoughts? [20:09] roaksoax, isn't that a known problem ^^ [20:13] kiko: no, but there amy be related to it updating the dns name instead ? [20:14] hmm [21:15] Bug #1549478 opened: DNS Forwarders not being added to bind configuration [21:43] roaksoax, is it possible to configure bond0 without configuring any subnet on it, and then have just vlan interfaces under? [21:59] nagyz: you should be able to have this for example: [21:59] bond0 fabric0 untagged subnet1 Unconfigured [21:59] bond0.1 fabric0 vlan1 subnet2 AutoAssign [22:07] generally or via the GUI as well? [22:08] ah I see I need to configure subnet1 on it - that's where the confusion is coming from [22:08] on a trunk interface it doesn't make sense for a bond interface to belong to a subnet [22:08] especially if I want subnet1 to be on bond0.1 [22:16] nagyz: yeah [22:17] nagyz: right, but it will still be unconfigured, but you could also put bond0.1 [22:17] nagyz: if you set the subnet as unconfigured, it does not allow you to create a bond ? [22:17] err a vlan interface [22:20] hmm, I could have sworn that it doesn't let me [22:20] but just tried it, and it works. [22:21] one more question: in some cases we see a node that we wanted to redeploy (after a release) booting into the previously deployed OS instead of doing a proper PXE boot [22:21] and going into the BIOS tells us that the order didn't change [22:21] should release reset the bios disk order? [22:30] also it seems kinda redundant to select a VLAN then a subnet [22:30] while the subnet is tied to vlan... [22:31] right now I cannot put the same interface both untagged and tagged to different nodes, even though they would be the same subnet, can I, roaksoax ? [23:50] help [23:50] The nodes leaving behind in the power error state after the nodes are commissioned and changed to ready state.