Daprd 1.5.0 does not respond to shutdown API command down in a timely fashion on Windows.

[badgeratu]

In what area(s)?

/area runtime

What version of Dapr?

1.5.0

Expected Behavior

On Windows 10 sending the POST HTTP shutdown command to /v1.0/shutdown should shut down the daprd.exe process in around 5 seconds.

Actual Behavior

The sidecar does not shut down in a timely fashion unless the same command is re-executed and then it forcibly shuts down after 5 seconds.

Steps to Reproduce the Problem

With a clean installation of dapr 1.5.0 using dapr init on Windows 10, run the following command (assuming port 80 is an open HTTP port on the machine):

daprd.exe --app-id demo --dapr-http-port 3500 --app-port 80 --log-level debug

Once it has started make an empty HTTP POST call to http://localhost:3500/v1.0/shutdown, wait 60 seconds then send the same call again to actually shut it down. The following log messages should be observed (with comments added prefixed by ##):

## Dapr Started
time="2021-11-19T17:31:28.1915287Z" level=info msg="starting Dapr Runtime -- version 1.5.0 -- commit 83fe579f5dc93bef1ce3b464d3167a225a3aff3a" app_id=demo instance=LONL30559 scope=dapr.runtime type=log ver=1.5.0
time="2021-11-19T17:31:28.2964495Z" level=info msg="dapr initialized. Status: Running. Init Elapsed 95.86970000000001ms" app_id=demo instance=LONL30559 scope=dapr.runtime type=log ver=1.5.0

## Sent http://localhost:3500/v1.0/shutdown
time="2021-11-19T17:31:40.2926015Z" level=info msg="Shutting down actor" app_id=demo instance=LONL30559 scope=dapr.runtime type=log ver=1.5.0
time="2021-11-19T17:31:40.2931354Z" level=info msg="dapr shutting down." app_id=demo instance=LONL30559 scope=dapr.runtime type=log ver=1.5.0
time="2021-11-19T17:31:40.2931354Z" level=info msg="Stopping Dapr APIs" app_id=demo instance=LONL30559 scope=dapr.runtime type=log ver=1.5.0

## 63 seconds later no change, repeated send of http://localhost:3500/v1.0/shutdown
time="2021-11-19T17:32:43.883986Z" level=info msg="Shutting down actor" app_id=demo instance=LONL30559 scope=dapr.runtime type=log ver=1.5.0
time="2021-11-19T17:32:43.8845826Z" level=info msg="dapr shutting down." app_id=demo instance=LONL30559 scope=dapr.runtime type=log ver=1.5.0
time="2021-11-19T17:32:43.8845826Z" level=info msg="Stopping Dapr APIs" app_id=demo instance=LONL30559 scope=dapr.runtime type=log ver=1.5.0
time="2021-11-19T17:32:43.9086103Z" level=info msg="Waiting 5s to finish outstanding operations" app_id=demo instance=LONL30559 scope=dapr.runtime type=log ver=1.5.0
time="2021-11-19T17:32:43.9086103Z" level=info msg="Waiting 5s to finish outstanding operations" app_id=demo instance=LONL30559 scope=dapr.runtime type=log ver=1.5.0
time="2021-11-19T17:32:48.9189706Z" level=info msg="Shutting down all components" app_id=demo instance=LONL30559 scope=dapr.runtime type=log ver=1.5.0

## Dapr has now closed successfully

When doing exactly the same approach using dapr 1.4.3, the second POST call is not required and the following log messages are observed:

## Dapr Started
time="2021-11-19T18:00:21.2112666Z" level=info msg="starting Dapr Runtime -- version 1.4.3 -- commit a8ee30180e1183e2a2e4d00c283448af6d73d0d0" app_id=demo instance=LONL30559 scope=dapr.runtime type=log ver=1.4.3
time="2021-11-19T18:00:21.3082046Z" level=info msg="dapr initialized. Status: Running. Init Elapsed 86.937ms" app_id=demo instance=LONL30559 scope=dapr.runtime type=log ver=1.4.3

## Sent http://localhost:3500/v1.0/shutdown
time="2021-11-19T18:00:25.1623406Z" level=info msg="Shutting down actor" app_id=demo instance=LONL30559 scope=dapr.runtime type=log ver=1.4.3
time="2021-11-19T18:00:25.1628614Z" level=info msg="dapr shutting down. Waiting 5s to finish outstanding operations" app_id=demo instance=LONL30559 scope=dapr.runtime type=log ver=1.4.3
time="2021-11-19T18:00:30.1674763Z" level=info msg="Shutting down all components" app_id=demo instance=LONL30559 scope=dapr.runtime type=log ver=1.4.3

## Dapr has now closed successfully

Release Note

RELEASE NOTE: FIX Bug in runtime. daprd does not respond to shutdown API command down in a timely fashion.

[shivamkm07]

/assign

[shivamkm07]

The issue stemmed from fasthttp not closing idle connections on Shutdown earlier. However, the same issue has been fixed and merged in upstream(valyala/fasthttp#1155). So just upgrading the dependency to latest commit(valyala/fasthttp@4517204) works.