=== zerick_ is now known as zerick | ||
=== kadams54 is now known as kadams54-away | ||
=== zz_CyberJacob is now known as CyberJacob | ||
=== CyberJacob is now known as zz_CyberJacob | ||
bloodearnest | it there a bug open for better formatting of the stdouts of a multi-unit juju run? Because it's unreadable at the minute | 09:47 |
---|---|---|
lazyPower | bloodearnest: not that i found on a quick search | 09:53 |
bloodearnest | lazyPower, yeah me neither, just wanted to check here before I filed one | 09:58 |
lazyPower | bloodearnest: you made mention you wanted to use the slides i dropped on speakerdeck. do you want the impress template? | 10:22 |
bloodearnest | lazyPower, yeah, that'd be sweet, thanks | 10:25 |
lazyPower | https://www.dropbox.com/s/pxapq90ma326nw3/service-orchestration.odp?dl=0 | 10:25 |
bloodearnest | lazyPower, sounds like you had fun :) | 10:25 |
lazyPower | bloodearnest: it was intense. one man show, 8 speaking events, 3 weeks notice. I was dancing, non stop :) | 10:25 |
lazyPower | well i had prior notice, but nothing was confirmed prior to 3 weeks before the conf. | 10:26 |
hazmat | lazyPower, intense | 12:20 |
lazyPower | hazmat: \o/ | 13:08 |
lazyPower | hazmat: signs of life! i thought you were going to be MIA for a while | 13:09 |
hazmat | lazyPower, i am for most of the day, restricted networks | 13:09 |
lazyPower | oi | 13:09 |
lazyPower | hazmat: did you have a chance to follow any of the slide work i published? | 13:11 |
lazyPower | i'm curious to get your take on what i put out there | 13:11 |
hazmat | lazyPower, checking | 13:12 |
hazmat | lazyPower, pretty solid | 13:15 |
lazyPower | hazmat: thanks for the review. Means quite a bit that you stamp it with approval :D | 13:16 |
* lazyPower hat tips | 13:16 | |
jcastro | hatch, ping | 14:44 |
hatch | jcastro: hey | 14:45 |
jcastro | hey so check this out | 14:45 |
jcastro | https://jujucharms.com/ceph/ | 14:45 |
jcastro | one revision behind, but james and crew pushed a new version on thursday | 14:45 |
hatch | did their version pass proof? | 14:45 |
marcoceppi-sast | https://store.juju.ubuntu.com/charm-info?charms=cs:trusty/ceph | 14:45 |
jcastro | aha! | 14:47 |
hatch | :) | 14:47 |
jcastro | E: Unknown relation field in relation nrpe-external-master - (gets) | 14:47 |
jcastro | yeah, got it, thanks! | 14:47 |
hatch | jcastro: np - we should probably have some indicator somewhere of something somehow for this :) | 14:47 |
* hatch intentionally leaves that very open ended :D | 14:48 | |
hatch | jcastro: typically if your charm doesn't show up in under an hour something went wrong :) it's on a 30min loop so depending on when you catch that loop... | 14:50 |
jcastro | ack | 14:50 |
=== Guest17528 is now known as rcj | ||
jcastro | hatch, would it make sense to autofile a bug on the charm if it fails? | 14:51 |
=== rcj is now known as Guest81467 | ||
hatch | jcastro: tbh I'd like to see some kind of a status page somewhere where you could check | 14:52 |
jcastro | well, not for me | 14:52 |
jcastro | for people who push and have no idea why it didn't show up | 14:52 |
hatch | right - it would be in the docs "go here for charm ingestion status" | 14:52 |
hatch | that could even possibly be linked by the page on jujucharms.com | 14:53 |
hatch | I'm totally just throwing out an idea here | 14:53 |
jcastro | iirc the original plan was to have the "status lights" on the charm's actual page | 14:54 |
hatch | yeah I think that filing a bug would be hard - I'm not sure that the proper people would see it for promulgated charms | 14:55 |
hatch | I don't think I see bugs filed for Ghost if it's on the promulgated charm? | 14:55 |
jcastro | hatch, found the problem, charm tools is returning 0 | 14:56 |
jcastro | so the ceph guys lint check, it's just returning(!) zero | 14:56 |
lazyPower | hatch: couple things about that | 14:56 |
lazyPower | 1) we already have the review queue - why not attach diagnostic messages like that to the review item? | 14:56 |
hazinhell | jcastro: the new store should have some events support .. ie if you use juju publish it should give you feedback | 14:56 |
lazyPower | 2) Why am i checking a page when you have the maintainer field in metadata and can contact the maintainer? | 14:56 |
jcastro | hazinhell, oh awesome, that sounds great. | 14:57 |
hatch | lazyPower: people can push to their own namespace (it'll still fail if proof fails) and it doesn't touch the review queue | 14:57 |
hatch | and the maintainer field isn't always accurate | 14:57 |
hazinhell | jcastro: its been supported for a few years fwiw (re events and juju publish for store feedback).. haven't tried it with the new store impl that's extant but it should work | 14:58 |
jcastro | right so instead we just say `juju publish fail, please run charm proof and fix your stuff | 14:58 |
=== cmars` is now known as cmars | ||
=== kadams54 is now known as kadams54-away | ||
hatch | is it normal for the debug-hooks command to not dump you into the hooks context on the 'start' hook? | 17:49 |
hazinhell | hatch: yes it dumps you into a empty window, and the others pop up in response to hooks executing, but if there is no active hook you won't be in a hook context | 17:50 |
hatch | ahh that's probably what's happening | 17:50 |
hatch | I wish I didn't have to spam debug-hooks after deploy | 17:50 |
hazinhell | hatch: there's some arcana involved but if you want to poke around at the unit context, you can use juju-run on the unit to examine its state. | 17:51 |
hazinhell | hatch: cory_fu made some nice debugging tools to avoid having to do the double poke (debug and resolved --retry) | 17:51 |
hatch | is there a reason why we don't say 'ok you asked for debug-hook, but the unit isn't up yet, so we'll wait till it is' | 17:51 |
hazinhell | hatch: no, there is no reason.. feel free to file it as a cli ux bug, there's a few like that | 17:53 |
hatch | Will do! thanks | 17:54 |
hatch | looks like there is already a bug :) https://bugs.launchpad.net/juju-core/+bug/1278831 | 17:57 |
mup | Bug #1278831: debugging first run of install hook is not straight forward <debug-hooks> <juju-core:Triaged> <https://launchpad.net/bugs/1278831> | 17:57 |
hatch | now to figure out how to use tmux in a tmux :) | 17:58 |
hazinhell | hatch: its easy.. rebind control key on the outer one ;-) | 18:09 |
hazinhell | hatch: that requires forethought.. the other way if they match on controls is to double hit the control char | 18:10 |
hatch | yeah there was definitely no forethought here :P | 18:16 |
=== mmcc` is now known as mmcc | ||
=== roadmr is now known as roadmr_afk | ||
=== roadmr_afk is now known as roadmr | ||
=== kadams54 is now known as kadams54-away | ||
dbart | mbruzek: ping | 22:12 |
mbruzek | hello dbart | 22:12 |
mbruzek | What can I do for you? | 22:12 |
dbart | mbruzek: hey there! I was working with lazyPower on the MariaDB charm | 22:13 |
mbruzek | dbart: Yes I heard that | 22:13 |
dbart | we had an issue with the charm on P8, but I think I've just solved it | 22:13 |
mbruzek | dbart: OK great news! | 22:14 |
dbart | MariaDB on P8 is built with IBM's Advanced Toolchain, and without the runtime package, MariaDB won't install | 22:14 |
dbart | so the issue was how to get the runtime | 22:14 |
mbruzek | dbart: "was" so you solved it? | 22:15 |
dbart | so what I've done is added the package to our MariaDB repository, so now when the charm goes to install mariadb-server it sees and pulls in the runtime package | 22:15 |
mbruzek | dbart: what is the package name? | 22:15 |
dbart | advance-toolchain-at8.0-runtime | 22:16 |
dbart | our packages (correctly) depend on it, but on the test box lazyPower was testing on the package wasn't in any of the configured repos, so apt-get failed | 22:17 |
mbruzek | $ apt-cache search advance-toolchain-at8.0-runtime | 22:17 |
mbruzek | I see no results for that dbart when I search on a power 8 system. | 22:17 |
dbart | exactly, it's generally found in a separate IBM repository | 22:18 |
dbart | so by adding it to our P8 repo when you try to install mariadb-server apt can find it | 22:18 |
mbruzek | dbart: This conversation seems familiar. Have we talked about this before? Does MariaDB need to be built with that package? | 22:19 |
dbart | IBM really wants us to build with it, it's not absoultely required, but performance is better when it's used | 22:19 |
mbruzek | dbart: It is my understanding that the advanced toolchain was needed for people who build with older kernels, and this was a way to get the new compiler to older level of kernels. My understanding is if you use modern kernels and compilers they are often newer than advanced toolchain. | 22:20 |
mbruzek | dbart: If the performance is better then I am not arguing | 22:20 |
mbruzek | dbart: From the conversations I have had it was obsolete and no advantage. | 22:21 |
mbruzek | to have the toolchain. | 22:21 |
dbart | ok, I don't know about all that, I just know that our devs are still building with it... | 22:21 |
dbart | I could ask them about it | 22:21 |
kwmonroe | negative mbruzek - AT brings cpu flags to the table that gcc hasn't brought yet | 22:25 |
mbruzek | dbart: I don't want to conflate the issue, if you have a fix that is great! I just was telling you what my understanding was of the advanced toolchain. Because we (Canoncial) asked about using advanced toolchain for other instances and I seem to remember that it was not needed. | 22:25 |
mbruzek | kwmonroe: ahh very good, thank you for correcting me. | 22:25 |
dbart | ok, understood :) | 22:25 |
kwmonroe | event based branching comes to mind.. as a power8 cpu feature that's not in gcc, but is available for AT | 22:25 |
kwmonroe | hardware transactional memory is another.. why buy a p8 if you're not gonna STEP UP TO THE POWAHHHHH?!?! | 22:26 |
dbart | mbruzek: so just a few minutes ago, with the runtime package in the repo, I was able to install MariaDB on the test instance lazyPower gave me access to | 22:27 |
mbruzek | dbart: it seems my memory is not as good as it used to be. Disregard my earlier statements about AT. | 22:27 |
dbart | :) | 22:27 |
mbruzek | dbart: how can I help? | 22:28 |
dbart | I'm going to be going afk now, and I don't know what else there is that needs to be done... lazyPower was working on testing the charm, but hit the repo issue | 22:28 |
dbart | so now that the repo issue is solved (and MariaDB can actually be installed) I assume testing can be resumed, but I don't know what lazyPower was going to do next on that | 22:29 |
mbruzek | dbart: what is your branch? | 22:30 |
dbart | https://code.launchpad.net/~dbart/charms/trusty/mariadb/trunk | 22:32 |
mbruzek | dbart so let me see if I understand, your install hook now installs the toolchain that was needed for mariadb to pass on ppc64le? | 22:33 |
dbart | there's an issue with the amulet tests (they don't account for the enterprise instructions in the README) | 22:33 |
mbruzek | dbart can you give me a hint? | 22:34 |
dbart | you basically need to create a ent.yaml file that has the correct information in order for the enterprise repo to work (which is where the P8 packages are) | 22:35 |
dbart | I don't know how lazyPower was doing the tests (since the amulet tests, out of the box, don't work) | 22:35 |
mbruzek | dbart: passing the amulet tests are something that are required for a charm to go into the recommended section of the charm store. | 22:36 |
mbruzek | dbart: Do you know what is needed to make them work? | 22:36 |
dbart | yes, so those need to be fixed, that's probably the next step | 22:36 |
kwmonroe | i know mbruzek.. the ./tests/10-deploy test is trying to install the "normal" maria packages first. that adds a repo and does an apt-get install mariadb. that's great, except the repo doesn't have a ppc64le arch.. so 10-deploy fails right off the bat when testing locally. | 22:37 |
dbart | kwmonroe: yup, that's it in a nutshell | 22:38 |
kwmonroe | on an arch that *does* exist in the mariadb repo, the next setp in 10-deploy is to "upgrade" the repo to use the secret enterprise url. but running the test natively on power, we can't get past step 1 to get to step 2. | 22:38 |
dbart | by using an ent.yaml file and passing that to deploy, you can set it up right of the bat | 22:38 |
kwmonroe | yup - just like the readme tells ya to do ;) | 22:39 |
kwmonroe | so mbruzek, the problem is making the amulet test do what the readme says. | 22:39 |
dbart | we really need a better solution for those that want to use the Enterprise repo, but what's outlined in the readme is all I've got at the moment | 22:40 |
dbart | lazyPower was working on updating the tests I think, but I don't know how far he got | 22:40 |
mbruzek | dbart: Can you refactor the amulet test to do that? I haven't seen or used mariadb in quite some time, a bit out of the loop | 22:40 |
kwmonroe | i think it gets even stickier dbart.. if you want to test the enterprise repo (which is the only option for p8), you need to expose a user:password in the amulet test. | 22:40 |
dbart | yup, that's the other issue | 22:40 |
dbart | I've been complaining to our team about removing the login requirement (nothing but trouble IMO), but as of today it's still needed | 22:42 |
dbart | tomorrow I can look into refactoring the test | 22:42 |
kwmonroe | dbart: is there perhaps an x-day trial user/pass? | 22:42 |
dbart | perhaps, I'd need to look into it | 22:42 |
dbart | anyway, I've got to go now, but I'll work on the tests (and see if lazyPower got anywhere with them) tomorrow | 22:44 |
dbart | thanks all | 22:44 |
kwmonroe | sure dbart - i'll help you manana in case you're sick of mbruzek | 22:44 |
kwmonroe | ;) | 22:44 |
dbart | thanks! | 22:44 |
dbart | :) | 22:44 |
mbruzek | dbart lazyPower is traveling across the ocean so I doubt he will be working on them | 22:44 |
dbart | oh yeah, he mentioned that... :-) | 22:45 |
mbruzek | dbart see you tomorrow let kwmonroe know if you need help refactoring the tests | 22:45 |
* mbruzek ducks | 22:46 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!