| timClicks | what's the easiest way to debug the apiserver? I want to know what's on the other side of the facade call | 03:09 |
|---|---|---|
| timClicks | ended up using this: juju debug-log -m controller --replay --include-module juju | 03:39 |
| babbageclunk | timClicks: for more detail you can turn logging in the apiserver up to trace using `juju model-config -m controller logging-config="juju.apiserver=TRACE"` | 03:59 |
| timClicks | anastasiamac: I think that this change might be harder than expected | 04:20 |
| timClicks | the code is all fine on the server-side, but the client-side API is lossy | 04:21 |
| timClicks | (c *Client) FindTools(majorVersion, minorVersion int, series, arch, agentStream string) (result params.FindToolsResult, err error) | 04:21 |
| timClicks | we loose the alpha/beta/rc and build number when we look for tools | 04:21 |
| timClicks | because c.AgentStream is "", the server defaults to released | 04:22 |
| hpidcock | If I'm adding a new field to a facade request params struct, should I bump the facade version? | 04:58 |
| babbageclunk | hpidcock: yes, particularly for a client facade (where there are other clients than just the juju CLI) | 05:00 |
| timClicks | hpidcock: in particular, the juju gui and pylibjuju also rely on that version number | 05:01 |
| timClicks | babbageclunk: btw your panic debugging method is pretty effective | 05:04 |
| hpidcock | +1 panic is great for debugging | 05:04 |
| babbageclunk | cheers! even more fun than printf debugging | 05:10 |
| hpidcock | more fun because it yields better results ;) | 05:15 |
| babbageclunk | who doesn't love a stack trace? | 05:16 |
| anastasiamac | timClicks: m happy to discuss further ur monday since m guessing u r eod now | 05:52 |
| anastasiamac | at eow even :D | 05:53 |
| hml | manadart: there’s something odd about what we’re doing with providerID for spaces. will look more. | 14:22 |
| hml | manadart: we’re adding spaces with an undefined name and -1 providerID. | 14:23 |
| rick_h | hml: so I think that's the MAAS "undefined" space | 14:25 |
| manadart | hml: Got a ref in code? We have the default space + MAAS undefined space. | 14:25 |
| manadart | What rick_h said. | 14:25 |
| rick_h | hml: manadart just for reference https://bugs.launchpad.net/juju/+bug/1785314 | 14:26 |
| hml | manadart: no code ref, i was looking at the DB | 14:26 |
| mup_ | Bug #1785314: maas spaces are mis-translated <track> <juju:Incomplete> <MAAS:Incomplete> <MAAS 2.3:Incomplete> <MAAS 2.4:Incomplete> <https://launchpad.net/bugs/1785314> | 14:26 |
| hml | rick_h: is the “undefined” space the default or something? | 14:28 |
| hml | there can be only 1? | 14:28 |
| rick_h | hml: yea, I think so. I have to refresh my brain | 14:28 |
| rick_h | hml: +1 to there being only 1 | 14:28 |
| rick_h | highlander space | 14:28 |
| hml | :-D | 14:28 |
| hml | rick_h: time to play queen | 14:29 |
| rick_h | wfm! | 14:29 |
| manadart | rick_h hml: I am leaning towards ignoring "undefined" when loading spaces and making the juju default space the unifying concept for the no/default/undefined/whocares space. | 14:33 |
| pmatulis | is not providing accurate information by default for the 'controllers' command really a good thing (need --refresh option to do that)? don't people always want to ensure they are getting accurate information? | 18:36 |
| rick_h | pmatulis: the issue is that controllers pulls a list and that info doesn't change much so a cache is usually ok | 18:38 |
| rick_h | pmatulis: otherwise calling each one (and timing out on ones you can't reach for say behind a vpn/etc) makes controllers risky to use | 18:38 |
| pmatulis | rick_h, if a client has a controller registered shouldn't we assume the client can reach it? anyway, implementing a short timeout doesn't seem too bad | 18:41 |
| === markthomas_ is now known as markthomas | ||
| pmatulis | anyway, i just upgraded a controller model and couldn't see the results. until i finally reached for the --refresh option. seemed like an odd UX moment is all | 18:42 |
| rick_h | pmatulis: no, again because it could be over a vpn, on a user's laptop and they can only reach it on the work network, etc | 18:44 |
| rick_h | pmatulis: yea, I wish we had a way to push from controller to clients like "you've just connected, I've got news for you" | 18:45 |
| rick_h | that would be a nicer way to handle stuff like that | 18:45 |
| pmatulis | rick_h, i see, ok | 18:51 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!