happy-oxygen-3414605/03/2023, 1:54 PM
without a project directory? Can I provide that command with what domain I want to authenticate with (app.garden as an example) instead of it figuring that out based on the project config?
icy-furniture-1751605/04/2023, 7:25 AM
folder when switching between envs to avoid deploy failures. is that something you do as well, or am I taking placebo? 🙂
ancient-diamond-8001105/04/2023, 9:34 AM
purple-oyster-2892705/04/2023, 9:45 AM
swift-garage-6118005/04/2023, 12:12 PM
big-spring-1494505/04/2023, 12:49 PM
garden run workflow
) can be quite confusing. E.g. with the following config,
fails with a very cryptic error message, and
kind: Project name: repro environments: - name: default providers: - name: exec initScript: exit 1 --- kind: Workflow name: test steps: - script: echo FOO name: setup
swift-garage-6118005/04/2023, 2:36 PM
in Bonsai. Feels more consistent with the rest of the CLI palette, and
just strikes me as inelegant.
chilly-gigabyte-8385305/04/2023, 3:45 PM
kind-france-8977705/04/2023, 8:26 PM
brief-restaurant-6367905/05/2023, 2:23 PM
instance again and again in the ws server instead of doing
This means that all the subscribers get added to the same instance. Now, when e.g. streaming logs, the subscriber callback gets called for each log event on all subscribers. So e.g. a log from the
command = new CommandClass()
service gets called on
et al (assuming the vote example project). This means the logs viewer will print the same log line for each subscribing service. I'm not entirely sure why this worked on 0.12 tbh but it's certainly an issue on 0.13. My first reaction would be to initialise the command per request as opposed to pre-initialising, but not sure how that jives with other invariants. We've been pre-initialising commands in the
file for forever. A quick fix would be to handle this in the callback and only send a message for the services the callback cares about, but that feels pretty hacky and doesn't generalise to other commands. Any thoughts? (cc @alert-helicopter-61082 , since you've done some work on this)
big-spring-1494505/05/2023, 3:49 PM
swift-garage-6118005/06/2023, 11:53 AM
option for the
command if we've got the
options? I think it might make more sense to just implement these three options (which put the command into monitor/persistent mode) and see if we still feel like we want the
option. That said, I'd be happy to add the
option if we feel strongly about including it. Just feels slightly unnecessary at a glance (i.e. it seems the user would always want to access the persistent behaviour in the context of log streaming, syncing and port-forwarding, or some combination of those).
swift-garage-6118005/06/2023, 12:29 PM
option to the
command on a branch I'm working on, and the output looks something like this. That is, the deploy log outputs look exactly like in the
command, which I think makes sense—it's minimal and doesn't waste horizontal space. Any objections or suggestions around this?
chilly-waitress-6259205/08/2023, 1:56 PM
is returning 502 bad gateway.
ancient-diamond-8001105/08/2023, 2:11 PM
helper in the testing utils - i noticed we don't have a similar helper for
, is there some good pattern in some test file that i could borrow? (basically, run a command, make sure we print out warnings in the garden log, but don't expect it to throw) EDIT: aha, found an example https://github.com/garden-io/garden/blob/92566652eed9e92313f256b245655df1c3b6ac0e/core/test/unit/src/commands/get/get-status.ts#L162-L191
ancient-diamond-8001105/08/2023, 4:42 PM
CI step at some point. the current implementation of the link checking is causing quite a bunch of spurious failures
swift-garage-6118005/09/2023, 1:37 PM
polite-fountain-2801005/09/2023, 3:22 PM
brief-restaurant-6367905/10/2023, 6:06 PM
(I know that's a module, just copied from existing config, but I figure it'll work the same with actions.) Previously I'd just run
kind: Module name: local type: exec local: true include:  services: - name: local deployCommand:  devMode: command: ["npm", "run", "dev"]
, and that deploys the stack, starts sync mode were applicable, and starts the local process. Or specifically
garden deploy --dev
garden deploy local --dev
flag be used here since obviously there's no sync going on? Or do we need to add special primitives around local processes (which is really not ideal imo)?
swift-garage-6118005/10/2023, 8:48 PM
purple-oyster-2892705/11/2023, 9:37 AM
via node/kubernetes-client ?
kubectl logs -f
swift-garage-6118005/11/2023, 9:48 AM
class. We use this in our logs follower, and add some logic around it (to deduplicate lines, handle timeouts and reconnects etc.).
alert-helicopter-6108205/11/2023, 10:29 AM
quaint-dress-83105/12/2023, 9:17 AM
brief-restaurant-6367905/12/2023, 9:26 AM
commands some weeks back on the K8s plugin. So it's
etc. Should we just remove these? Obviously a breaking change but I don't think they'll be missed, assuming we also add a proper
garden plugin kubernetes sync start
garden sync status