=== CyberJacob is now known as CyberJacob|Away [02:21] anybody have a few minutes to point me in the right direction? [02:22] jason___: depends. where are you trying to get to? [02:23] trying to get farther then a pxe boot and get them to install and work on an HP c7000 [02:48] still digging through all the info i can find :P [03:11] bigjools: do you remember what we decided about the API problem of "PUT one field, reset all the others"? Do we simply accept that clients have to read and re-PUT the full object? [03:21] jtv: I don't remember if we even decided anything [03:21] I think PUTing the whole thing makes sense [03:21] For some value of sense. [03:21] since you can't create that way [03:21] but I expect we have examples where that's not the case [04:23] jtv: if you want a break from coding I have a lovely branch up for review :) [04:24] Coding? I wish. [05:35] I knew that word once [06:35] jtv: you scheduled the IPv6 meeting for FRiday next week :) [06:35] 344 [06:35] err [06:36] Argh! [06:36] * jtv heads back to Google [06:37] And the wrong timezone, too... === CyberJacob|Away is now known as CyberJacob === CyberJacob is now known as CyberJacob|Away === jpds_ is now known as jpds === jpds is now known as Guest15370 === Guest15370 is now known as jpds_ === jamespag` is now known as jamespage [11:30] morning folks [11:35] Hi onicrom [11:53] fingers crossed i can get this 'custom' partition working in curtin [12:25] ok i seem to have found whats going on [12:29] it looks like the default for my system is pt_mbr [12:29] which has a bug [12:35] it also looks like the params passed to pt_mbr are ignored. [12:35] also a bug [12:43] ok submitted a bug [12:48] /query/win 16 [12:48] err === jfarschman is now known as MilesDenver [14:03] Does anyone have time for a shortish review? https://code.launchpad.net/~allenap/maas/rpc-alt-cluster-rpc-fixture/+merge/230109 [14:54] allenap: done [14:54] blake_r: Thanks! [15:01] allenap: https://code.launchpad.net/~blake-rouse/maas/add-largefile-migration-fix/+merge/230129 [15:02] allenap: could I get a really quick review, for my stupidity! [15:02] blake_r: Sure. [15:02] blake_r: Is IntegerField really not enough? Woo. [15:03] allenap: yeah it is, but might not be if file is larger than 4tb, that should never happen! [15:03] allenap: sorry I mean large than 2gb [15:03] allenap: you set it to approved? not you approving it [15:03] blake_r: Sorry! Fixed. [15:04] allenap: thanks [15:13] allenap: my branch may have a huge diff but it's actually really really simple: moved a bunch of functions from provisioningserver.utils to provisioningserver.utils.twisted. [15:22] ok i need to speak to the guy who wrote curtin :) [15:22] i have hacked the /usr/lib/curtin/helper/common script 6 ways from sunday [15:23] i can see that my chages are being executed, but the installed system always always creates a 2tb root filesystem [15:23] I think that's smoser. [15:26] blake_r, did you say you figured out how to make curtin sing & dance? [15:27] jtv: I stay far away from the bash scripting that smoser wrote! [15:27] Good man. But _someone_ said they figured it out and it wasn't so hard once you cracked the secret. [15:27] jhobbs maybe? [15:27] jtv: i wrote the uefi partitioner in the bash [15:27] jtv: i cried most of the time, lol [15:28] (Still think they should have named the GNU shell "gnash") [15:28] that's the flash player [15:28] It is _now_, yes. But at the time there wasn't one. :) [15:28] onicrom: if you have a paste of your bash script I can take a look [15:29] blake im updating the bug [15:30] blake_r: https://bugs.launchpad.net/maas/+bug/1354426 [15:30] Ubuntu bug 1354426 in MAAS "curtin creates a 2TB root fs/partition by default" [Undecided,New] [15:31] thanks ubot5 :) [15:32] onicrom: you want the partition bigger than 2tb? [15:32] onicrom: you cannot create a partition larger than 2tb on mbr partition layout [15:32] onicrom: you need to use a gpt partition layout [15:34] i know [15:34] i dont want to [15:34] i want to create a 50Gb root partition [15:35] the default option in common is to use mbr, i tried to switch this to gpt and modify the size values there but that just errored completely the pt_gpt function is broken [15:43] onicrom: are you sure that '1,6490,L' is correct? [15:43] based on the sfdisk man page yes [15:43] ive tried using cylendars sectors etc [15:43] same thing [15:44] based on some math that is the correct number of cyls for 50gb [15:44] or ~ anyway im not that picky [15:54] looks like its only 50gb partition, thats what you want, you just need another line for the other partition [15:54] what other partition? [15:57] no matter what i do something is creating a 2TB partition [16:02] so weird doesnt matter what i execute something creates a 2TB partition... i wonder... let me try a brand new machine without any existing table maybe the bit that wipes the existing partitions is broken :) === matsubara is now known as matsubara-lunch [16:17] blake_r: have you run the 1.6 (or 1.5, 1.4 etc.) tests lately? I'm seeing a test failure in trunk that's related to commissioning timeouts, and it looks like it might be timezone-sensitive. [16:18] You're in a negative UTC offset now, right? [16:18] jtv: unit tests? i am running trunk [16:18] jtv: never had an issue [16:19] Right, unit tests — specifically, src/maasserver/tests/test_api_commissioning.py:TestCommissioningTimeout [16:19] jtv: i am -5 [16:19] jtv: never seen a failure there [16:19] Yes, if this is the timezones, then trunk should pass for you — but maybe you'd get the converse bug in 1.6 or older. [16:19] jtv: i will try 1.6 [16:19] Where the tests creates a node that started commissioning an hour and one second ago, but it doesn't time out yet. [16:19] Thanks! [16:20] You'd have to run "make && ./bin/test.maas src/maasserver/tests/test_api_commissioning.py:TestCommissioningTimeout" [16:21] (I filed this as bug 1354511...) [16:21] bug 1354511 in MAAS "check_commissioning fails test: test_check_with_commissioning_but_not_expired_node" [Critical,Triaged] https://launchpad.net/bugs/1354511 [16:21] jtv: all 3 passed for me [16:21] Grrrrr [16:21] Thanks. [16:21] Your welcome. [16:28] blake_r: I think I'm one step closer! My databases in 1.6 and trunk seem to be running at different timezone offsets... [16:28] It may have been the travel that caused it. [16:29] jtv: isn't the database created before each test run? [16:29] jtv: its different than make syncdb [16:29] Yes, but maybe the local database _cluster_ that lives inside the branch still had a remnant of UTC+1... [16:30] (Wow, I don't know what type of animal I just slapped off my leg but it had a lot more rump than a mosquito and it spilled a lot of blood...) [16:30] (Mine, of course!) [16:31] got a lot of mosquitos where your at? [16:32] Ohhh yes. [16:33] Not a lot of bites so far, I think. These are not those little Euro bugs — you feel the sting WHILE they're biting you. [16:36] Damn. [16:37] That's a funny change in habits I had to go through: when I felt a mosquito bite, I used to say to myself "oh well, it's already happened, all I can do now is try not to scratch it." So I reacted slowly on purpose. [16:38] Here, when you feel the bite, you need to hurry so you can still get rid of the bastard! [16:38] * jtv looks accusingly at 3 geckos overhead that failed to catch this one [16:38] jtv: where are you? the jungle? lol [16:38] Pretty much. [16:41] is there a way to make the fastpath installer the default? [16:45] I am trying to get some modifications to dhcpd.conf.template to take effect, but saving the cluster controller does not update /etc/maas/dhcpd.conf. [16:45] ^ Any suggestions? [16:46] (I am running trusty) [16:47] onicrom: it's the default now. [16:48] ctlaugh: try saving a cluster controller interface rather than just the cluster controller. You may have to make a small change just to convince it. [16:49] jtv, not in the version thats installed with trusty [16:49] onicrom: then marking a node for the fast-path installer sets a tag. [16:49] jtv: I tried that as well -- still not getting updated [16:49] ls -al [16:49] jtv is there a way to set that tag on all nodes upon creation? [16:49] * ctlaugh typed in the wrong window by mistake [16:50] onicrom: I think there is, but I haven't done it myself — set an xpath expression for the tag that simply always matches. [16:50] onicrom: then for any new node, the system will see if it matches the xpath expression and if so, give it that tag. [16:50] jtv: Is there any way outside of the web ui (like a command line) to force an update? [16:51] Thinking... [16:51] You'd have to update a cluster interface field that actually affected the dhcp settings, such as increasing the dynamic IP range by 1 address (and then taking it back again if you care enough). [16:59] im beginning to think that the 'helpers' in curtin really dont do anything [17:05] jtv: That's what I did. I changed on end of the address range. [17:06] Hmmm... that ought to restart the dhcp server. [17:06] Assuming it's an interface where maas manages dhcp. [17:07] jtv: Not only is it not picking up my changes (from the template), it's also not getting the range change from what I modified in the UI. [17:07] How can I verify that it thinks it's managing dhcp on that interface? [17:07] Something I also recently did was change the IP address of the maas server. [17:08] Should be simple: the cluster must have been accepted into the maas, and its management setting should be either "DHCP" or "DHCP and DNS." [17:09] UI shows accepted, and interface shows "Manage DHCP and DNS" [17:09] Owww, the mechanism that we use to carry messages from the region controller to the cluster controller, RabbitMQ, doesn't take too kindly to changing addresses. [17:09] That would explain it, I think. [17:10] (And yes, we're now in the process of getting rid of it!) [17:11] I ran dpkg-reconfigure on both maas-cluster-controller and maas-region-controller to fix the addresses, and think I rebooted after doing that (but possibly before) [17:11] Is there somewhere else that might need to be updated? [17:13] Try a "grep -rI $OLD_ADDRESS /etc/rabbit*" [17:14] jtv: no matches, and /etc/rabbitmq is empty [17:16] Oh, or /var/lib/rabbit* perhaps? [17:16] We may store that stuff in the maas /etc or /var trees... I remember running into this in the past and just grepping like mad. [17:16] And it did turn up rabbit files. [17:18] jtv: ok, after a reboot and modifying the interface again, it worked. [17:18] \o/ [17:19] apparently, I changed the ip address in /etc/interfaces, rebooted, then ran dpkg-reconfigure [17:19] I should have restarted things after the reconfigure as well [17:20] jtv: thank you for your help [17:20] Glad it worked out. [17:20] As always, "have you tried turning it off and on again?" :) [17:21] Yes :) === matsubara-lunch is now known as matsubara [17:43] Hi -- has anyone noticed/reported that vms with virtio disks do not correctly register all their disks in MAAS? Is that something that would be fixed if I reported a bug? :) [17:54] can someone running 1.5.2 verify what is on line 225 [17:54] of /usr/lib/curtin/helpers/common === roadmr is now known as roadmr_afk === CyberJacob|Away is now known as CyberJacob [19:09] if I could get a review on both of these: https://code.launchpad.net/~blake-rouse/maas/add-bootresource-models/+merge/230161 and https://code.launchpad.net/~blake-rouse/maas/add-bootresource-models/+merge/230161 [19:10] i was nice enough to split them up for the reviewer! [19:10] :) [19:10] crap same link twice, sorry: https://code.launchpad.net/~blake-rouse/maas/bootresource-model-properties/+merge/230164 === cmagina_ is now known as cmagina [19:22] blake_r: having a look, and liking the documentation so far... [19:43] blake_r: reviewed one. [19:53] can someone point me to the scripts/source around cloud-init ? === roadmr_afk is now known as roadmr [19:53] it's doing something to the partition table... i ran a set -x on the curtin partition scripts and nother in there is doing this [21:33] can someone point me to the code that does the initial boot after the 'start' is completed === wmf is now known as thewmf [21:53] howdy maas people [21:53] I'm trying to get multiple NICs to work with curtin but the blog post doesn't make sense to me [21:54] anybody interested? [22:01] im having a fun time with curtin :) [22:04] so you're into S&M? [22:07] i wanna poke out my eyeballs [22:09] ok, that's the same as my experience [22:09] I actually have experience with d-i, but I don't think it's working [22:15] im trying to change the default partition size [22:15] its nearly impossible and im hacking code [22:15] I saw that thread on the list. not promising [22:16] I hate the d-i partitioner [22:32] onicrom, still hitting your head against the way with curtin? [22:32] ;) [22:32] dood [22:32] im dying here [22:33] im about to change the grub to drop me to a shell [22:33] onicrom, the only code that I have modified for curtin so far has been to add the ability to add a separate boot partition, which is only supported for gpt currently [22:33] and howd that work? [22:33] i switched to gpt [22:33] good, its in trunk now [22:34] what you will need to do is you will need to make new block_meta commands [22:34] to be able to set the size [22:34] can you paste me like to the code? [22:34] take a look at curtin/curtin/commands/block_meta.py [22:35] also.... the part that is confusing me is im able to get it to create the partition layout i want [22:35] this is the code that calls the "partition" script which then calls the "common" script [22:35] and when helpers/common runs and exits sgdisk -p /dev/sda [22:35] prints out what i want [22:35] then it reboots and its gone and changed sda1 to be the full isize of the disk [22:35] are you modifying templates? [22:36] where are said templates? [22:36] let me re-phrase my question [22:36] How are you calling curtin? [22:36] or are you just hacking the code in curtin? [22:37] hacking the code in curtin [22:38] well ... im editing the common shell script which is exected by the partition shell script which is executed by block_meta [22:38] okay so yeah you can go about it that way too [22:39] have you seen my long ass bug? [22:39] no, do you have a link? [22:39] https://bugs.launchpad.net/maas/+bug/1354426 [22:40] Ubuntu bug 1354426 in MAAS "curtin creates a 2TB root fs/partition by default" [Undecided,New] [22:40] the most relevant bit where i have it outputting exactly what im doing [22:40] then the install finishes and the damn gremlins did somethin [22:41] * newell takes a look [22:42] ha...the gremlins [22:45] is there any way you can convert to the newest code? [22:45] sure might need help :) [22:45] okay no problem [22:45] I can help [22:46] $ sudo apt-get install bzr [22:46] you need bzr to check out the source [22:46] ok [22:46] once you have bzr you can create a branch [22:46] give me 2 mins [22:46] need to potty break [22:46] $ bzr branch lp:curtin curtin [22:46] that will make a directory called curtin for you [22:47] $ cd curtin [22:47] $ ./tools/build-deb -uc -us [22:47] The above command will build all the *.deb packages for you that you need [22:47] then you can install them doing: [22:47] $ sudo dpkg -i *.deb [22:53] done [22:54] onicrom, everything worked? [22:54] for install that is [22:55] yes im going to rebuild one box [22:55] just to see [22:55] onicrom, so now you know how to take the source and make a package out of it [22:56] so, you can hack it up, then install it etc [22:56] thats pretty sweet [22:56] the results you were mentioning were for hardware correct? [22:56] you mentioned etemplates [22:56] not the devel stuff? [22:56] what do you mean hardware results/ [22:57] I mentioned to you in the email how if you follow doc/devel/README.txt it will show you how to develop for curtin with qemu images [22:57] I was wondering if the results you mentioned in the bug report are for that (qemu) or on actual hardware? [22:57] ohhhh that yes hardware [22:58] everything im doing is on baremetal [22:58] k [22:58] i have 20 servers im trying to get ready for openstack :) [22:58] nice :) [22:59] so let me know what issues you run into later on but at least you have the most current code installed [22:59] i only have ipmi access via serial over lan... and the debian installer hoses the network for ipmi somehow (the nic is shared between the ipmi and the os) [22:59] so im forced to use curtin [22:59] as it does not do such network fuckery [23:00] do you know what happens with regards to partitions/disks upon first boot after the curtin bit is run? [23:02] you mean after curtin installs? [23:02] correct [23:03] ok so that didnt work [23:03] i have no /boot [23:03] you need to use the simple-boot command [23:03] not the default simple [23:03] currently it is dependent on which architecture you have [23:05] how do i select [23:07] change in block_meta or ? [23:08] onicrom, give me a second [23:08] for you [23:08] take 2! [23:13] onicrom, sorry about that [23:13] np i hard coded [23:13] take a look at doc/topics/overview.rst [23:13] boot=true [23:14] yeah you can do that too ;) [23:14] that will make you a 512MiB separate /boot partition on /dev/sda1 [23:15] and the rest of the disk will have / partition on /dev/sda2 [23:15] i changed that to +50G [23:15] i hope hope hope that works [23:15] $end is bad for me :) [23:15] yeah curtin could really use this feature [23:16] ok good news [23:16] no boot partition [23:16] BUT [23:16] root is 50gG!@!!!!!!@#$@%$@#% [23:16] how are you verifying the partitions? [23:17] sorry let me rephrase [23:17] /boot is there, but there is no fstab reference [23:17] that is because you hardcoded it [23:17] the fstab code is in block_meta.py [23:17] ah :) [23:17] yeah its a cluster fsck [23:18] onicrom, what you can do is you can hard code the default to get the fstab [23:19] instead of doing the hack in common [23:19] change... [23:19] * newell looks where it is [23:19] install.py at the top has block-meta using 'simple' [23:19] change that to 'simple-boot' [23:19] that or you can specify the partitioning command as the overview.rst shows [23:20] lemme try that one [23:20] i still need to hack common [23:21] yeah for the size part you will [23:21] --new "1::+1G" --typecode=1:8300 \ [23:21] --new "2::+50G" --typecode=2:8300 "$target" || [23:21] it be nice to pass those variables just lke passing simple-boot [23:21] file a bug report for the ability to create custom boot partitions (i.e. the size) [23:22] hell it would be simple to pass 1:size:mount and have common iterate over an array [23:22] as well as block_meta [23:23] yeah shouldn't be too hard [23:23] feel free to add it and contribute if you like :) [23:24] i would put some disclaimer... if you put in numbers here that do not match the disks available its your own damn fault [23:25] i mean right now it tries to be smart about the sizes and what not but i say let people shoot themselves in the foot :) [23:30] would paritioning_commands:./curtin/commands/block_meta.py: 'choices': ['raid0', SIMPLE, SIMPLE_BOOT]}), [23:30] inside curtin_preseed? === CyberJacob is now known as CyberJacob|Away [23:32] onicrom, did you take a look at overview.rst? [23:32] it shows how to do the partitioning commands [23:32] vi ues [23:32] 50_setup_raid: curtin disk-setup --all-disks raid0 / [23:32] yes [23:33] so that gives examples [23:33] change raid0 to SIMPLE_BOOT [23:33] no [23:33] you would do similar to what is install here [23:34] 00_something_blah: curtin block-meta simple-boot... [23:34] something along those lines [23:34] similar to install.py* [23:35] ahhh block_meta [23:35] I have never done it this way so not sure on the exact syntax [23:35] but I know install.py is doing this (those are the defaults at the top if nothing in the configuration sets it) [23:35] 'partitioning_commands': { [23:35] 'builtin': ['curtin', 'block-meta', 'simple']}, [23:36] there we go [23:36] paritioning_commands: 10_doshit: ['curtin', 'block-meta', 'simple-boot'] [23:37] i really need to pick up python :) [23:37] yeap [23:37] so where you going to put that last line? [23:38] /etc/maas/preseeds/curtin_userdata [23:38] yeap ;) [23:38] i threw it after early_commands and before late_commands [23:38] go ahead and give that a go and let me know how it works [23:38] correct [23:38] that is what overview.rst says at least [23:38] its running now :) [23:39] ok cross your fingers the install is going [23:40] the part that was confusing me earlier... i did a set -x in the common script saw exactly what was being run [23:40] and i was at a complete loss as to how the partition size was changing [23:40] cool, let me know how it goes [23:42] question about python with the file.py and file.pyc [23:42] will it exec the pyc? so the contents of the py could have been one way and the pyc did something compeltely different? [23:43] pyc is executed by if there are any modifications to the original py before running it will re-compile the byte code for the pyc [23:43] ok. [23:43] so you should be good, as long as you saved the file ;) [23:43] also my new machine now has a 2.7TB fs .... [23:44] is that what you wanted? [23:44] it juuust had a 50G wtf [23:44] nope [23:44] common has the code changes the only difference this time was boot=true was set to false [23:45] you changed the curtin_userdata? [23:45] correct [23:45] did it at least create the separate /boot partition? [23:45] but i would think even without that change it would have executed common which would have done: --new "1::+50G" --typecode=1:8300 "$target" || [23:46] what does sudo mount say when you login to the node [23:46] ? [23:46] what common file are you editing? [23:46] did you re-create the package again? [23:47] also for maas, when you change the curtin_userdata...which file are you changing? [23:48] all those things matter [23:48] mount [23:48] /dev/sda1 on / type ext4 (rw) [23:49] and the normal tmpfs stuff [23:49] k, what about the other questions? [23:49] /etc/maas/preseeds/curtin_userdata [23:49] editing: /usr/lib/curtin/helpers/common [23:52] maybe print to the console to make sure you are executing what you think you should be in common? [23:53] im going to set -x that bitch! [23:53] :) [23:54] I have never used that, how does it work? [23:54] i was very close to adding a /bin/bash to the grub line it enters to drop me to a prompt before the first boot [23:54] lol [23:55] set -x in bash is essentially debug mode, it will show you everything its doing [23:55] cool never used it but playing with it now [23:56] so why are you setting up different partitions, is that something you think you would use often? [23:57] i want to use lvm for most of my storage so i can do lvm snapshots for my lxc containers [23:57] if i also want to setup drdb replication [23:57] linux has always has /boot as a non-journal'd filesystem for safety [23:58] so i want that and generally i throw everything in lvm volumes so that if i need to grow them later its super easy [23:58] ok so something broke :) [23:59] what did set -x tell you? [23:59] https://paste.tinyw.in/index.php/view/11650792 [23:59] it didnt get that far this time..