Fix HealthManager

[thinker0]

• Fix legacy of HealthManager
• Fix Sfl4jBridge (Unsupported)
• Add healthcheck API for heron-tracker, heron-ui
• Fix heron-tracker, heron-ui

[nicknezis]

What was the error encountered that required the change from float to string in the timeline object? I ask because there were various issues with the intermingling of the types. I was hoping we could keep it consistent, but recognize there might be use cases I didn't account for with the prior testing. So I'm interested to better understand what this PR fixes, and also does the timeline in the UI still work? If so, then I think I'm ok with this PR.

[surahman]

I ask because there were various issues with the intermingling of the types. I was hoping we could keep it consistent, but recognize there might be use cases I didn't account for with the prior testing.

I approved because Python typing will, rather annoyingly, allow you to achieve what follows. I agree with the need for consistency.

float_Val: float = 9.19
print(float_Val, type(float_Val))

str_Val: str = float_Val
print(str_Val, type(str_Val))

str_Val = 919
print(str_Val, type(str_Val))

"""
9.19 <class 'float'>
9.19 <class 'float'>
919 <class 'int'>
"""

[thinker0]

incubator-heron/heron/tools/tracker/tests/python/query_operator_unittest.py

Lines 41 to 57 in 700125f

	return MetricsTimeline(
	starttime = 40,
	endtime = 360,
	component = "a",
	timeline = {
	"c": {
	"b": {
	40: "1.0",
	100: "1.0",
	160: "1.0",
	220: "1.0",
	280: "1.0",
	340: "1.0"
	}
	}
	},
	)

In the internal UnitTest, it seems to be a String.

[nicknezis]

Should the unit test be updated to be a float instead of string? I did not catch that.

When I started working on the last PR, it was originally an int which I think was incorrect. str might be valid option. I think I tried to make it work, but we had enough other issues in the PR. I trust your opinion on str vs float.

We should make sure we are setting the type consistently. Should this line of code also be updated to be a str?

Edit: Also, does this other line, which was unchanged in the prior PR, help the choice to use str? It seems the code was converting to float which wouldn't make sense with my use of float.

[thinker0]

humm. Conflit is occurring in several places ㅠ.ㅠ

[thinker0]

I fixed it on my CentOS 7.

[thinker0]

%  ./docker/scripts/test-unittest.sh centos7 0.20.5
================================================================================
INFO: Elapsed time: 1889.214s, Critical Path: 207.25s
INFO: 3826 processes: 1175 internal, 2651 local.
INFO: Build completed successfully, 3826 total actions
Test cases: finished with 1189 passing and 0 failing out of 1189 test cases

Executed 247 out of 247 tests: 247 tests pass.
INFO: Build completed successfully, 3826 total actions
Cleaning up scratch dir

[thinker0]

% ./docker/scripts/test-unittest.sh rocky8 0.20.5
================================================================================
INFO: Elapsed time: 1835.725s, Critical Path: 246.64s
INFO: 3826 processes: 1175 internal, 2651 local.
INFO: Build completed successfully, 3826 total actions
Test cases: finished with 1189 passing and 0 failing out of 1189 test cases

Executed 247 out of 247 tests: 247 tests pass.
INFO: Build completed successfully, 3826 total actions
Cleaning up scratch dir

[thinker0]

Should the unit test be updated to be a float instead of string? I did not catch that.

Nothing

We should make sure we are setting the type consistently. Should this line of code also be updated to be a str?

I don't think you need to change this part.
When 0.20.5 centos binary was applied to both tracker and ui in production, it worked in the UI regardless of this part.

[surahman]

There was a failing test for killing topologies but I have seen that fail intermittently before. I did restart the CI pipeline.

Is the downgrade of the dev-toolset related to an unsupported library causing issues?

I do not see any issues with using the C++1y flag instead of C++14 because the other OSs have their version of GCC and G++ kept up to date. This means the flag will indicate compiling against the final C++14 standard on those OSs. It will, however, compile using the draft standard on Centos7 (GCC 4.8.x) but given the build completes successfully I believe this should be okay.

I will request @nicknezis to have a look at the tracker change because he has spent a lot of time in it recently.

[thinker0]

Fixed some functions.

• When Topology is re-executed (updated), it is giving a message that some '/stats' data of Tmanager is missing. In this case, since some data was missing in the tracker, logic to check this part was added, and if there is no data in this part in the UI, a function to check the value was added to the part where the screen was drawn.
• Topology Zookeeper Update sync error #3820
  -- https://github.com/apache/incubator-heron/pull/3819/files#diff-479615841054520a32d26e4a6857c244f63a704b6355292f35d7a96969f3156dL463
  -- https://github.com/apache/incubator-heron/pull/3819/files#diff-479615841054520a32d26e4a6857c244f63a704b6355292f35d7a96969f3156dL480
  -- https://github.com/apache/incubator-heron/pull/3819/files#diff-84d3871b76c715140b59ab14bd4b10351c8bd997c34590cf63e0aef53a4e36bcR997

Received response from Tmanager: Unknown component: __stmgr__
Received response from Tmanager: Metrics not available for component 'xxxxxx'

• Fixed because the current status could not be displayed in the UI
  -- https://github.com/apache/incubator-heron/pull/3819/files#diff-29e7366069fde10ea907fccddae5506e19a2d8dcaa5f515ce417fec4957e10ffR69

• Added because there is no Instance Healthcheck API in Tracker and UI.
  -- https://github.com/apache/incubator-heron/pull/3819/files#diff-555d27350199002b67424764ea3ec5f2078e9d04c6d8067b1cb6c8a60b8fc70dR132
  -- https://github.com/apache/incubator-heron/pull/3819/files#diff-761812c31ae5763145f0e4fc31e09a80d398d85a306498386d42132d6a3d40cbR617

[thinker0]

There was a failing test for killing topologies but I have seen that fail intermittently before. I did restart the CI pipeline.

Is the downgrade of the dev-toolset related to an unsupported library causing issues?

I do not see any issues with using the C++1y flag instead of C++14 because the other OSs have their version of GCC and G++ kept up to date. This means the flag will indicate compiling against the final C++14 standard on those OSs. It will, however, compile using the draft standard on Centos7 (GCC 4.8.x) but given the build completes successfully I believe this should be okay.

I will request @nicknezis to have a look at the tracker change because he has spent a lot of time in it recently.

This was judged to be a cause other than the cause of gcc, so it was restored to its original state.