hi team, we're having issues with metric collection on the recent flyteadmin. Telegraf starts throwing lots of these after just a day or two of uptime and some metrics disappear from the dashboard:

Error in plugin [inputs.prometheus]: error reading body: net/http: request canceled (Client.Timeout exceeded while reading body)

original settings were:

interval = "10s"
response_timeout = "3s"

I changed these to 30s and 15s and it seems to be ok for now. Do you have a recommendation on that these should be set to?

Hi, everyone! Any tips to help debugging Helm errors when trying to start the new sandbox lite? I can only see

Starting Docker daemon...
Done.
Starting k3s cluster...
Done.
Deploying Flyte...
"flyteorg" has been added to your repositories
Release "flyte" does not exist. Installing it now.
Error: timed out waiting for the condition

@katrina @Prafulla Mahindrakar @Haytham Abuelfutuh I updated to Flyte

1.0.1

and now workflows executed with

flyte-cli execute-launch-plan

(with flytekit 0.26.0) lose their k8s service account that was declared with

flyte-cli register-files --kubernetes-service-account

. When I execute the same workflow from Flyte Console the k8s service account is fine.

I am running a bunch of subworkflows from my parent workflow. I want to run them in serial (instead of in parallel). Can I chain them together like I can chain together normal Flyte tasks (i.e. https://docs.flyte.org/projects/cookbook/en/latest/auto/core/control_flow/chain_tasks.html)?

hi, we used to rely on this file in the sandbox image https://github.com/flyteorg/flyte/blob/master/deployment/sandbox/flyte_generated.yaml with

dind

tag but it seems I can’t find it anymore, any idea?

Hey, everyone. Did the in-container path for secrets change for 1.x.x? I can't access my secrets anymore, and I can't find the directory

/etc/flyte/secrets

either

hi, has this metric been renamed recently? (@katrina)

flyte-admin-task.execution.manager-task.executions.terminated.counter

Hi! I was looking at this metric from propeller called

grpc_client_handled_total

but I only see requests to the AdminService, is there a way to see requests to datacatalog as well?

Hi everyone. I'm new to using flyte. I'm using wsl and have installed flytectl using curl. How do I address "flytectl command not found?

Hi Flyers, Join us in welcoming our Flyte OS Community Sync guest speaker, Freenome's @jeev, who will tell us more about how Flyte can benefit the biotech industry, in Taking Flyte at Freenome. Tomorrow 5/17, 9am PT Zoom Link: https://www.addevent.com/event/EA7823958 See you there! Flyte Team

Plus how does someone deal missing modules such while writing python scripts using flyte functionalities?

Hello all. I'm attempting to use flytectl 0.5.25 on Centos7. I am trying to create a project. I keep getting the attached SIGSEGV, similar to (but not exactly) the issue here: https://github.com/flyteorg/flyte/issues/1320. What am I doing wrong here?

Also join the Flyte Community Sync tomorrow for a demo of pyflyte run (script mode) by Union.ai's @Eduardo Apolinario (eapolinario) 5/17, 9am PT Zoom Link: https://addevent.com/event/EA7823958

Have a Flyte question? Want to chat? Drop in to our Weekly on Wednesdays Flyte office hours: - Red Eye Flyte with @Haytham Abuelfutuh, 7:00am PT (Zoom link) -✨*New*✨Office Hours with @katrina, 1:30pm PT (Zoom link) - Late Night MLOps with @Ketan (kumare3), 9:00pm PT (Zoom link) Check out the #announcements channel for reminders!

Hey #flyte team, I want to use

FlyteRemote

to execute a workflow launch plan where one of the parameters is my custom class (assume that I have a good reason why I want to do this). For example, I have:

@dataclass_json
@dataclass
class WrappedList:
    files: str

and the workflow

@flytekit.workflow
def LocalSlam(input_list: WrappedList)

and that I want to use

FlyteRemote

to execute this worklow like

lp = remote.fetch_launch_plan(name=workflow_name, version=workflow_version)
e = remote.execute(lp, inputs=inputs)

Hi! Is there a way to get current image FQN (in case of a remote execution) from within a task? Specifically I would be interested in having that information in

PythonTask.pre_execute()

, so

user_params: ExecutionParameters

would also be available if that helps.

Hi all! 👋 I've started exploring flyte recently but I don't seem to be able to run workflows on my sandbox environment. I've created a sandbox environment using the command

flytectl sandbox start

and exported the env vars for kube and flyte configuration. I'm trying to run the hello world example from flytesnacks and I've created a virtual env using poetry, in which I installed pyflyte (1.0.2). I can run it locally no problem, but when I add the

remote

flag, I get a connection reset by peer error. Any ideas what is wrong on my setup? I'm on WSL2

Hi Flyers, Many thanks to folks who attended yesterday's Flyte Community Sync, and special thanks to our speakers! Recordings have been published on Flyte's YouTube channel: • Community Update - @katrina () • Script Mode - pyflyte run - @Eduardo Apolinario (eapolinario) () • Taking Flyte at Freenome - @jeev () Feel free to check out the Meeting Notes, and mark your calendars for our next sync on Tuesday May 31st, featuring MethaneSAT! Flyte Team

Hi all 👋 The new release of FlyteConsole v.1.1.0 is now available. 🎉 It includes: • Refactor to monorepo structure to simplify future support for plugin system • StructuredDataSet Input/Output type support • Updated TaskDetails, with link to TaskDetails page and ability to collapse parts of the schema • Better representation for map-task in Graph view (more to come…) • Fix to ensure that API requests for users with different admin/console domain will work as expected • Add interruptible override to launch forms • Number of cosmetic fixes Full list of changes could be found in #flyte-console channel

I don't know exactly where to put it but would it be possible to push a fix for this issue? https://github.com/flyteorg/flyte/issues/2444 We had to downgrade as half of our data scientists couldn't log-in because of that bug 😅

Hey, everyone! I'm looking at the new parts of the documentation and wondering if the serialization/registration process changed on 1.x.x. The Getting Started example executes a workflow locally using the default Docker image in the sandbox server, right? But I thought that the code was required to be inside the container so that Flyte could run it. Does pyflyte run an implicit fast serialization? If not, what is it doing? I'm curious because our team wants to use Flyte on production, but I fear that we'll have a lot of very similar Docker images (since we mostly use the same libraries). Having "base" images that we could use on simple tasks/projects would be a life-saver.

Hello! We have a webapi plugin and would like to control the rate of how often propeller calls the

Status()

hook [interface explained]. We have noticed that this is very much depending on load. During low (normal) load

Status()

is called multiple times per second. The

Get()

hook is only executed once every 30s, so we consider any rate of

Status()

higher than 15s to be a waste. The webapi

PluginConfig

contains configurations for a

ReadRateLimiter

and

WriteRateLimiter

[github] which I am unable to find being used in either propeller nor plugin machinery. We are not experiencing any throttling from the external API, only that the

Status()

rate is very high. Anyone have any pointers where to go from here?

I am looking for a way to pass Numpy Arrays (ndarray) and PyTorch/Tensorflow Tensors as Flyte Task input/output. I haven’t come across any example yet. I’m aware of the native support for Dataframes. It seems inefficient to convert ndarray/Tensors back and forth using Dataframes. How are folks handling this?

When using the API, what is the format for specifying filters. ex: I want to fetch a project with a certain name. The API docs all say:

Indicates a list of filters passed as string. More info on constructing filters : +optional.

but I don't see anywhere where it explains the structure of the filters

I'm getting a typing complaint while writing an integration test for a workflow: 🧵

Hey! I'm trying to set up schedules for my workflows (LaunchPlan.get_or_create -> remote.register_launch_plan -> flyte update launchplan --activate in cli). The schedule does show up in Flyte console (eg. as "Every hour"), but the workflow is never executed. In the scheduler pod there's an error message

rpc error: code = InvalidArgument desc = missing key in inputs

. Any ideas what could be wrong? More code in the thread.

Hi Flyers, Flyte 1.0 was featured in GitHub's Release Radar blog this month - thanks to our awesome community! https://github.blog/2022-05-20-release-radar-apr-2022/

Is there a way to override/add a particular option when calling the

pyflyte register

command? I.E. I’m adding a new field in flytekit and in the Container IDL proto to add an architecture override, is there a way to force that during the register for all workflows in a github repo, rather than needing to modify all of the workflows one by one? @Anmol had mentioned this to me before, but we couldn’t find any docs about it.

Hi everyone! Is there any plan to add a way to define task dependencies just using a requiriment.txt file or dependency name, as it is in Airflow's PythonVirtualenvOperator?