=== wallyworld_ is now known as wallyworld | ||
=== CyberJacob|Away is now known as CyberJacob | ||
=== vladk|offline is now known as vladk | ||
rvba | bigjools: Do you really need to resort to raw SQL in get_hostname_ip_mapping? | 07:20 |
---|---|---|
bigjools | hoist. petard. etc | 07:21 |
bigjools | rvba: yes. | 07:21 |
bigjools | same as the other same-named func does | 07:21 |
rvba | Ah, right, I think I see why now… | 07:23 |
bigjools | me too | 07:24 |
bigjools | debugging it made it clear, it's not the same dict | 07:24 |
rvba | bigjools: I /think/ Django grew support for "Distinct ON". | 07:24 |
bigjools | well, still not sure why it's lazy, but uh | 07:25 |
bigjools | oh maybe it's a memory thing | 07:27 |
bigjools | rvba: "I'd rather properly test the behavior" | 07:30 |
rvba | bigjools: in this branch you've added StaticIPAddressManager.get_hostname_ip_mapping but it's not used anywhere yet… do you have a follow-up branch that makes use of it? | 07:30 |
bigjools | testing behaviour: that's what I am avoiding | 07:30 |
bigjools | it's slow | 07:30 |
bigjools | yes I am using it locally but struggling to get tests to pass | 07:31 |
rvba | Yeah, I meant end-to-end behavior. For the reason I stated. But I'm okay with mocking, it's just not as reliable (and that's why you missed the fact that the old get_hostname_ip_mapping wasn't doing the right thing). | 07:31 |
bigjools | these tests are a nightmare | 07:32 |
bigjools | they are running the same code twice because of signals | 07:32 |
rvba | The ones in test_dns? | 07:33 |
bigjools | what wasn't it doing? | 07:33 |
bigjools | yes test_dns | 07:33 |
bigjools | make_nodegroup etc etc causes signals | 07:33 |
rvba | Your tests aren't exercising the code up to the point where get_hostname_ip_mapping is called. | 07:34 |
=== CyberJacob is now known as CyberJacob|Away | ||
bigjools | so the same dns code is run | 07:34 |
bigjools | rvba: yes I know, that is the intention | 07:34 |
rvba | Ah, okay, cool then. | 07:34 |
bigjools | it's tested elsewhere | 07:34 |
bigjools | well, ideally in these :) | 07:34 |
bigjools | rvba: every time create_managed_nodegroup() is called in the test it fires a dns job through a signal | 07:35 |
rvba | bigjools: yeah, but until we call 'self.patch(settings, 'DNS_CONNECT', True)' that doesn't trigger a real task. | 07:35 |
rvba | bigjools: you'll see that 'self.patch(settings, 'DNS_CONNECT', True)' is called *after* all the setup is done. | 07:36 |
bigjools | true, but, still feels like one of the main reasons tests are slow | 07:36 |
rvba | It's possible. | 07:36 |
bigjools | they have gone massively slow on me lately, something has changed | 07:36 |
rvba | I didn't notice anything this dramatic on my side. | 07:37 |
bigjools | and we have not put enough importance on the fact that fast tests are a necessity | 07:37 |
rvba | I agree with that. | 07:37 |
bigjools | hurray a passing test | 07:39 |
* bigjools fixes others | 07:39 | |
bigjools | rvba: very glad you made create_nodegroup_with_lease() | 07:41 |
bigjools | easy to change to create static ip instead and thusly fix many tests in one fell swoop | 07:41 |
rvba | Encapsulation FTW :) | 07:42 |
bigjools | wooo passing tests | 07:43 |
bigjools | I'll push a new rev in a sec | 07:43 |
bigjools | rvba: annoyingly there's a bi-dependency between node and dns now, so inline imports are needed :( | 07:45 |
bigjools | rvba: dns.py can't import from maasserver.models.staticipaddress | 07:45 |
=== roadmr is now known as roadmr_afk | ||
rvba | I see. | 07:45 |
bigjools | rvba: anyway new rev pushed, can you have a look please? I am going to eat, back in a little while | 07:46 |
rvba | bigjools: looking at the code, I think you're right and the only reason why lazydict was introduced was to make use of __slots__ and thus save memory space in case the mapping is huge. | 07:47 |
rvba | bigjools: sure, looking now. | 07:47 |
bigjools | yeah | 07:47 |
bigjools | bbiab | 07:47 |
=== mwhudson is now known as mwhudson- | ||
=== mwhudson is now known as mwhudson-bip | ||
bigjools | rvba: "deviation from the original design"... not as I understood it! We spoke yesterday and said that dynamic IPs do not need dns CNAME entries | 08:27 |
=== mwhudson-bip is now known as mwhudson | ||
=== roadmr_afk is now known as roadmr | ||
=== vladk is now known as vladk|offline | ||
=== mwhudson is now known as mwhudson-bip | ||
jtv | gmb: I remember seeing a bunch of "6" suffixes on keywords in an IPv6 sample config. | 09:48 |
gmb | jtv: Yeah, it really is pretty much that simple :) | 09:48 |
bigjools | Ran 2647 tests in 1889.564s | 09:49 |
gmb | (at least on my test network, on a small scale) | 09:49 |
bigjools | urgh | 09:49 |
jtv | Why, why, why does factory.make_node() perform two update queries on the node it just created, and then another query to load back the object!? | 09:49 |
gmb | bigjools: The v6 stuff couldn’t have increased runtimes could it? I.e. if there’s lots of “if IP in network” where network is a /64; that kind of thing? | 09:50 |
bigjools | doubt it | 09:50 |
gmb | Mind you we’re not doing it in many places yet. | 09:50 |
bigjools | exactly | 09:50 |
bigjools | profile. we need a profile | 09:50 |
bigjools | rvba: the branch is all done now, if you want to look again. All I did was fix a couple of failing tests and remove the dns update for new leases. | 09:57 |
* jtv longs for a utilities package | 09:58 | |
jtv | gmb: did you land your make-a-random-ipv6-address factory method? | 10:01 |
gmb | jtv: Yes. it’s factory.get_random_ipv6_address(). | 10:02 |
rvba | bigjools: LGTM | 10:02 |
jtv | gmb: wha!? Why not just make_ipv6_address()? | 10:05 |
jtv | Maybe I should have said... | 10:05 |
rvba | bigjools: are you witnessing a general slowdown or is it confined to a subset of the tests? | 10:05 |
bigjools | rvba: general | 10:05 |
bigjools | rvba: thanks, BTW | 10:06 |
rvba | The branch I described this morning is up for review: https://code.launchpad.net/~rvb/maas/utopic-syslinux/+merge/224182 | 10:17 |
=== roadmr is now known as roadmr_afk | ||
=== vladk|offline is now known as vladk | ||
jtv | rvba: blason added a file to the list of boot loader files, but it's present in syslinux 4 and 6 alike. | 11:04 |
jtv | poweroff.com, I think | 11:04 |
rvba | Yeah. I think we can live with a bit of duplication here. | 11:04 |
=== vladk is now known as vladk|offline | ||
=== vladk|offline is now known as vladk | ||
=== vladk is now known as vladk|offline | ||
=== vladk|offline is now known as vladk | ||
AskUbuntu | Unable to add nodes to ubuntu MAAS server on static vlan | http://askubuntu.com/q/487549 | 11:32 |
=== roadmr_afk is now known as roadmr | ||
=== roadmr is now known as roadmr_afk | ||
rvba | bigjools: gmb: turns out we *can* use omshell to query the leases. I just tested it and it works. | 13:00 |
gmb | rvba: Sweet. | 13:00 |
=== vladk is now known as vladk|offline | ||
=== jtv_ is now known as jtv | ||
=== roadmr_afk is now known as roadmr | ||
blake_r | jtv: "maasutils" package would be very nice | 13:10 |
blake_r | jtv: "ensure_dir" is not in maasserver, ugh! | 13:10 |
jtv | blake_r: yeah, so far we've mostly had to dump generic helpers in provisioningserver, because provisioningserver can't import from maasserver (but vice versa works). | 13:12 |
blake_r | jtv: oh yeah forgot, was thinking it was the other way around | 13:13 |
jtv | There shouldn't be a need to remember... "maasutils" | 13:14 |
=== vladk|offline is now known as vladk | ||
rvba | blake_r: btw, FYI using a symlink /syslinux/ at the root of the TFTP dir worked. | 13:20 |
blake_r | rvba: i was looking at that | 13:21 |
blake_r | rvba: so you created a symlink from current/syslinux -> /usr/lib/syslinux ? | 13:21 |
rvba | blake_r: exactly | 13:21 |
rvba | blake_r: but syslinux 4 and 6 seem to have different requirements for what *really* needs to be stored at the top level. | 13:22 |
blake_r | rvba: i thought pxelinux expected it all to be in tftproot | 13:22 |
blake_r | rvba: does it check both? | 13:22 |
rvba | blake_r: syslinux 6 definitely does | 13:22 |
blake_r | rvba: cool | 13:23 |
blake_r | rvba: that makes it easy | 13:23 |
rvba | Yeah | 13:23 |
=== jfarschman is now known as MilesDenver | ||
rvba | newell_: Hi there. I see you haven't had time to work on the MP I marked as "Needs fixing". I realize what I asked is probably a bit more work than you had anticipated because it implies dealing with forms. But in the long run, it's a worthy investment (for the various reasons I stated in my review); I'm happy to give you a hand with that if you want. | 14:14 |
jtv | rvba: lander's tests broke on your branch. :( | 14:35 |
jtv | The syslinux one. | 14:35 |
rvba | jtv: I saw that. The failure looks completely unrelated. | 14:36 |
jtv | Yeah. And I remember that was a function I disliked for some reason. | 14:36 |
jtv | I wonder if it's a function I messed with. | 14:36 |
rvba | jtv: it's a spurious failure. It passes locally from time to time. | 14:37 |
jtv | You mean it _fails_ locally from time to time, I hope? It's been passing for me. | 14:38 |
rvba | Not sure, I haven't run it a great number of times. | 14:39 |
rvba | But yeah, it seems success is more frequent than failure. | 14:40 |
jtv | I bet there's an undetected clash in networks or ranges. | 14:40 |
rvba | I'm going to investigate but I'd like to get my syslinux branch landed first. | 14:40 |
=== jfarschman is now known as MilesDenver | ||
newell_ | rvba, I haven't worked on that because I was told that there were some other items that were higher priority, so I have been working on those. | 14:44 |
newell_ | Yes, your help would be appreciated | 14:44 |
newell_ | Once I have finished the higher priority items that I am currently working on, I will ping you so you we can go over it. | 14:45 |
newell_ | Thanks for lending a helping hand. | 14:45 |
rvba | newell_: okay, cool. | 14:48 |
d_` | where in the source is the startup script for bootstrapping a node? | 15:03 |
d_` | based on packet captures on both hosts, they're making a bunch of DNS queries that are failing, but based on the hosts they're querying, I'm guessing some of them are supposed to fail? | 15:07 |
jtv | Who's up for a code review? https://code.launchpad.net/~jtv/maas/arp-on-failed-entries/+merge/224322 | 15:39 |
roaksoax | blake_r: ^^ | 15:42 |
blake_r | roaksoax: sure | 15:42 |
roaksoax | blake_r: thanks! | 15:43 |
blake_r | jtv: done! | 15:52 |
jtv | Thanks blake_r! | 15:52 |
=== vladk is now known as vladk|offline | ||
blake_r | jtv: np | 15:52 |
jtv | You say that, but actually it hasn't even been proven that np isn't the same thing as p. | 15:53 |
blake_r | Yeah that makes since! But I will keep saying "np"! | 15:54 |
rvba | blake_r: rarg, poweroff.com is poweroff.c32 in Utopic's syslinux. | 15:57 |
blake_r | roaksoax: it should have both | 15:57 |
blake_r | rvba: ^ | 15:58 |
blake_r | rvba: poweroff.com at /usr/lib/syslinux | 15:58 |
blake_r | rvba: from syslinux-common package | 15:58 |
rvba | That's not the case. I just installed syslinux-common and syslinux-dev. | 15:58 |
blake_r | rvba: ugh | 15:58 |
blake_r | rvba: http://packages.ubuntu.com/search?searchon=contents&keywords=poweroff.com&mode=exactfilename&suite=utopic&arch=any | 15:59 |
rvba | blake_r: http://paste.ubuntu.com/7695727/ | 16:00 |
blake_r | rvba: that is so wierd, guess the packages site is incorrect | 16:01 |
blake_r | rvba: guess need a symlink from poweroff.com -> poweroff.c32 | 16:01 |
rvba | blake_r: indeed. Maybe roaksoax will know how that could happen. | 16:01 |
rvba | blake_r: ewww | 16:01 |
blake_r | rvba: temp[late needs to work fro both trusty and utopic | 16:01 |
rvba | blake_r: I know. | 16:02 |
jtv | And precise? jtv interjects with a stupid look | 16:02 |
rvba | But I don't want to do a symlink if we can avoid it. | 16:02 |
rvba | blake_r: I wonder if, with the other symlink I just added it won't just work out of the box if we remove poweroff.com from the list of things we copy over. | 16:03 |
=== roadmr is now known as roadmr_afk | ||
rvba | blake_r: I guess it's worth a shot. I'll test that tomorrow. | 16:04 |
rvba | blake_r: are, I didn't realize the template was referencing the file directly… | 16:08 |
rvba | s/are/arg/ | 16:08 |
loki27_ | I have question about charms deployment and juju .. If i deploy the same charm , on different machine, (Let's say mysql and rabbitMQ) , how will the deployment work their relation, will the mysql instances be standalone , or are they going to replicate the same data trough all instances ? | 16:10 |
rvba | blake_r: isn't there a way to say: "poweroff.com or poweroff.c32"? | 16:10 |
jtv | loki27_: you may have more luck in #juju with questions specific to a Juju charm. | 16:11 |
loki27_ | haa great | 16:12 |
jtv | (Not saying nobody will know, but I think it's the best chance) | 16:12 |
jtv | gmb: looks like you put code in make_interface_settings to leave static_ip_range_{low,high} set to None if the random range has no room for them. | 16:13 |
jtv | Guess what happens when you enter None in the dict that goes into a POST? | 16:13 |
gmb | jtv: Not exactly; it was to allow for the old (dynamic only) behaviour to be maintained… I’m guessing that that doesn’t work properly :/ | 16:13 |
jtv | Right. It blows up, but only on rare runs. | 16:14 |
* jtv doesn't really know what gmb just said, but finds enough common ground to bluff his way through | 16:14 | |
jtv | These are the things that need documenting in code! | 16:15 |
gmb | I thought I had; my apologies :/ | 16:15 |
gmb | jtv: Can you paste me a failure? | 16:15 |
jtv | Urgh. I think I just lost that shell. Let me reproduce it... | 16:15 |
blake_r | rvba: I don't think so | 16:16 |
blake_r | rvba: I will look and see | 16:16 |
rvba | Okay, ta. | 16:17 |
blake_r | rvba: from what I can find, doesn't seem to be anything about handling a missing file or alternatives | 17:08 |
=== CyberJacob|Away is now known as CyberJacob | ||
lazypower | strange. When i remote into my VMAAS nodes and issue df -h i dont see a root parition listed | 18:06 |
lazypower | is this expected behavior? | 18:06 |
=== CyberJacob is now known as CyberJacob|Away | ||
d_` | lazypower: you don't see a root partition? how is your system operating? | 18:31 |
lazypower | d_`: my sentiments exactly | 18:33 |
lazypower | let me rebootstrap maas, spin up a machine with juju and pastebin the output for df -h | 18:34 |
lazypower | http://paste.ubuntu.com/7696549/ | 18:58 |
lazypower | d_`: sorry for the delay, but thats what i see out of df -h on my maas nodes | 18:58 |
lazypower | they all have 8gb disks allocated. This is a new one on me where df doesn't report all mounted filesystems. | 18:59 |
blake_r | lazypower: whats the content of /etc/fstab | 19:05 |
lazypower | blake_r: LABEL=cloudimg-rootfs / ext4 defaults 0 0 | 19:18 |
lazypower | that's the only line in there - so is this a byproduct of using some kind of overlayfs? | 19:18 |
blake_r | lazypower: that is correct | 19:28 |
blake_r | lazypower: i mean the corrent fstab | 19:28 |
lazypower | ok | 19:29 |
blake_r | lazypower: what does blkid, show | 19:29 |
lazypower | this appears to have changed recently - within the last 4 months. I recall a while back df-h showed the contents of / mounted and its disk space | 19:29 |
lazypower | blkid shows nothing, returns 0 | 19:29 |
lazypower | do i need to pass it an arg? | 19:29 |
blake_r | no | 19:29 |
blake_r | try "ls /dev | grep sd" | 19:30 |
lazypower | nothin | 19:30 |
blake_r | try | 19:31 |
blake_r | try "ls /dev | grep vd" | 19:31 |
lazypower | that yields 2 disks. vda device and vda1 | 19:31 |
blake_r | so your in a physical machine? | 19:31 |
lazypower | well, disk / partition | 19:31 |
blake_r | i mean virtual machine | 19:31 |
lazypower | i'm in a virtual machine. correct | 19:31 |
lazypower | i'm orchestrating VMAAS. my HOST (metal) is the region/cluster controller | 19:31 |
lazypower | i'm using kvm w/ about 13 nodes pre-allocated as my maas cluster. | 19:32 |
blake_r | so vda1 is your "/" | 19:32 |
blake_r | fdisk /dev/vda | 19:32 |
blake_r | 19:32 | |
blake_r | should show partition, correct? | 19:32 |
lazypower | correct | 19:32 |
blake_r | what version of maas? | 19:32 |
blake_r | apt-cache policy maas | 19:33 |
lazypower | Installed: 1.5.1+bzr2269-0ubuntu0.1 | 19:33 |
blake_r | can you view the curtin log, as the node is installing ubuntu? | 19:34 |
lazypower | i'm not sure whre that lives - i normally open up the VM from VMM and watch the logs there. | 19:34 |
blake_r | okay, do this | 19:34 |
blake_r | on the webui of maas, go to the settings page and set the global kernel parameters to this value | 19:35 |
blake_r | console=ttyS0 | 19:35 |
blake_r | then "start" a node | 19:35 |
blake_r | then on your host do "virsh console {name-of-vm}" | 19:35 |
blake_r | that will show the booting log and curtin output | 19:35 |
lazypower | right on, neat | 19:36 |
blake_r | you should do "virsh console", quickly | 19:36 |
lazypower | yeah i'm not seeing anything. should i see the TFTP output? | 19:36 |
lazypower | er, the pxe boot sequence | 19:37 |
lazypower | blake_r: ah, found the problem. I've got the curtin log now. | 19:41 |
blake_r | lazypower: perfect | 19:41 |
lazypower | blake_r: am i looking for something specific in here or do i need to pastebin this for you? | 19:47 |
blake_r | lazypower: oh thought you saw the problem | 19:48 |
blake_r | lazypower: yeah you can just pastebin the whole thing | 19:48 |
lazypower | [ 2.565478] EXT4-fs (sda1): mounted filesystem with ordered data mode. Opts: (null) | 19:48 |
lazypower | thats teh only thing i see pertinent to what i'm looking for... | 19:48 |
lazypower | ok let me re run this with output redirected to a logfile | 19:49 |
lazypower | blake_r: http://paste.ubuntu.com/7696786/ | 19:54 |
lazypower | wait, i see this here. init bottom says something about tmpfs as the overlayfs | 19:55 |
lazypower | wat | 19:55 |
blake_r | thats just for install | 19:56 |
blake_r | your system reboots | 19:57 |
lazypower | ah ok | 19:57 |
blake_r | does this new node have the same problem? | 19:57 |
lazypower | blake_r: yeah :( | 20:01 |
blake_r | lazypower: did you change the preseed in anyway | 20:02 |
lazypower | blake_r: nosir | 20:03 |
blake_r | lazypower: that is so wierd | 20:03 |
blake_r | lazypower: i would recommend re importing the boot images | 20:03 |
blake_r | lazypower: rm -rf /var/lib/maas/boot-resources/current/* | 20:03 |
blake_r | lazypower: then "maas-import-pxe-files" | 20:03 |
blake_r | lazypower: just to see if that is it | 20:04 |
lazypower | already on it. i wiped the entire directory though | 20:04 |
lazypower | /var/lib/maas/boot-resources | 20:04 |
blake_r | that should be okay | 20:04 |
lazypower | blake_r: yeah still seeing the same behavior after reimporting the pxe images | 20:47 |
blake_r | lazypower: you are deploying precise, can you try to deploy trusty | 20:48 |
lazypower | strange... my default series is set to trusty | 20:48 |
lazypower | ah wait, i'm wrong. | 20:49 |
lazypower | blake_r: mind = blown | 20:53 |
lazypower | works as expected with trusty | 20:53 |
blake_r | lazypower: strange precise should work | 20:53 |
blake_r | lazypower: alteast it is working now | 20:53 |
lazypower | i mean, it appears to work. the thing that got me was being unable to report disk usage | 20:54 |
lazypower | i'm deploying a mongodb cluster, and ran out of disk space thanks to preallloc - which is what prompted this discovery session | 20:54 |
blake_r | might ask in #ubuntu-server they might know an answer for precise | 20:56 |
lazypower | thanks for the troubleshooting tutorial blake_r. this was informative | 21:00 |
blake_r | lazypower: np | 21:00 |
lazypower | i thought i had a good grasp on maas until this :P | 21:01 |
lazypower | now i realize there's so much more | 21:01 |
=== CyberJacob|Away is now known as CyberJacob | ||
=== mwhudson-bip is now known as mwhudson | ||
=== CyberJacob is now known as CyberJacob|Away | ||
=== jfarschman is now known as MilesDenver | ||
=== jfarschman is now known as MilesDenver |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!