feat: Replace `sessions`, `kernels`'s `status_history`'s type `dict` with `list` #3201

jopemachine · 2024-12-05T04:48:08Z

Resolves #3200 (BA-20), Part of #3214 (BA-14).
Refs #412, follow-up to #480.

The current implementation saves only the most recent timestamp whenever status information in status_history is updated, and all previous information is deleted.

This PR prevents the loss of timestamp information by changing the data structure of sessions, kernels's status_history to List.

Previous format:

{
  "PENDING": "2024-01-01T00:00:05Z",
  "SCHEDULED": "2024-01-01T00:00:10Z",
  "PREPARING": "2024-01-01T00:00:12Z",
  "RESTARTING": "2024-01-01T00:30:00Z",
  "RUNNING": "2024-01-01T00:30:08Z",  // overwritten if restarted
}

New format:

[
  {"status": "PENDING", "timestamp": "2024-01-01T00:00:05Z"},
  {"status": "SCHEDULED", "timestamp": "2024-01-01T00:00:05Z"},
  {"status": "PREPARING", "timestamp": "2024-01-01T00:00:05Z"},
  {"status": "RUNNING", "timestamp": "2024-01-01T00:00:15Z"},  // preserved
  {"status": "RESTARTING", "timestamp": "2024-01-01T00:30:00Z"},
  {"status": "RUNNING", "timestamp": "2024-01-01T00:30:08Z"},
]

Checklist: (if applicable)

Milestone metadata specifying the target backport version
API server-client counterparts (e.g., manager API -> client SDK)

📚 Documentation preview 📚: https://sorna--3201.org.readthedocs.build/en/3201/

📚 Documentation preview 📚: https://sorna-ko--3201.org.readthedocs.build/ko/3201/

jopemachine · 2024-12-05T04:48:24Z

This stack of pull requests is managed by Graphite. Learn more about stacking.

jopemachine · 2024-12-05T06:23:01Z

@HyeockJinKim How can we specifically implement a length restriction?
We need to decide whether the length restriction should be configurable, whether older logs in the status_history_log list should simply be deleted when the limit is exceeded, and whether a warning log should be printed during the deletion process.

src/ai/backend/manager/models/utils.py

jopemachine · 2024-12-10T05:02:52Z

Broken CI will be fixed in #3235.

HyeockJinKim · 2024-12-17T01:31:28Z

src/ai/backend/client/cli/session/lifecycle.py

@@ -844,48 +846,63 @@ def logs(session_id: str, kernel: str | None) -> None:
            sys.exit(ExitCode.FAILURE)


-@session.command("status-history")
+@session.command()


I don't know much about session.command, but can you subtract the internal string "status-history"?

backend.ai/src/ai/backend/client/cli/session/lifecycle.py

Lines 59 to 61 in 42b4030

@main.group()

def session():

"""Set of compute session operations"""

Looking at the code, it seems that the decorator itself does not perform any operations on the arguments.
So, I think it would be better to leave the arguments empty, as is currently done in most of the other code.

I think this is using overload to overload the function for the last implementation, can you double check? @jopemachine

If no value is passed as an argument, the decorated function name is automatically used as the CLI command.
If we pass an arbitrary string, that string can be used as the CLI command instead of changing the function name.
So, let's leave the arguments empty here, just like with the other functions.

HyeockJinKim · 2024-12-17T01:38:00Z

src/ai/backend/manager/models/minilang/ordering.py

+                    case JSONArrayFieldItem(_col_name, _conditions, _key_name):
+                        # TODO: Implement this.
+                        pass
+                        # ...


Was it handled in a follow-up PR?

This code requires discussion before going ahead.
I created a Teams thread, but I haven’t received any responses yet.

Copilot reviewed 6 out of 21 changed files in this pull request and generated 1 comment.

Files not reviewed (15)

src/ai/backend/manager/api/schema.graphql: Language not supported
src/ai/backend/manager/models/minilang/queryfilter.py: Evaluated as low risk
src/ai/backend/client/cli/session/lifecycle.py: Evaluated as low risk
src/ai/backend/client/output/fields.py: Evaluated as low risk
src/ai/backend/client/utils.py: Evaluated as low risk
changes/3201.feature.md: Evaluated as low risk
src/ai/backend/manager/server.py: Evaluated as low risk
src/ai/backend/manager/scheduler/dispatcher.py: Evaluated as low risk
src/ai/backend/manager/registry.py: Evaluated as low risk
src/ai/backend/manager/models/utils.py: Evaluated as low risk
src/ai/backend/manager/models/session.py: Evaluated as low risk
src/ai/backend/manager/models/resource_usage.py: Evaluated as low risk
src/ai/backend/manager/api/resource.py: Evaluated as low risk
src/ai/backend/manager/models/gql_models/kernel.py: Evaluated as low risk
src/ai/backend/manager/models/minilang/ordering.py: Evaluated as low risk

src/ai/backend/manager/models/gql_models/session.py

…with `list`

Co-authored-by: octodog <[email protected]>

…201.feature.md Co-authored-by: octodog <[email protected]>

github-actions bot assigned jopemachine Dec 5, 2024

github-actions bot added area:docs Documentations comp:manager Related to Manager component comp:client Related to Client component comp:cli Related to CLI component require:db-migration Automatically set when alembic migrations are added or updated labels Dec 5, 2024

github-actions bot added the size:XL 500~ LoC label Dec 5, 2024

jopemachine mentioned this pull request Dec 5, 2024

feat: Replace sessions, kernels's status_history's type map with list #2113

Closed

2 tasks

jopemachine added this to the 24.12 milestone Dec 5, 2024

jopemachine added type:feature Add new features type:refactor Refactor codes or add tests. labels Dec 5, 2024

jopemachine requested a review from HyeockJinKim December 5, 2024 06:03

jopemachine marked this pull request as ready for review December 5, 2024 06:03

This was referenced Dec 5, 2024

feat: Add dummy_kernels table for testing sql_json_merge #3204

Open

feat: session status-history handler #1116

Closed

This was referenced Dec 5, 2024

feat: Replace vfolder's status_history's type dict with list #3205

Open

Add missing status_history GQL field #3214

Open

jopemachine commented Dec 6, 2024

View reviewed changes

src/ai/backend/manager/models/utils.py Outdated Show resolved Hide resolved

jopemachine commented Dec 6, 2024

View reviewed changes

src/ai/backend/manager/models/utils.py Show resolved Hide resolved

jopemachine force-pushed the topic/12-05-feat_replace_sessions_kernels_s_status_history_s_type_dict_with_list_ branch 4 times, most recently from 656df34 to dceabcc Compare December 10, 2024 00:38

jopemachine force-pushed the topic/12-05-feat_replace_sessions_kernels_s_status_history_s_type_dict_with_list_ branch from 850165f to 98c21c0 Compare December 13, 2024 00:58

HyeockJinKim reviewed Dec 17, 2024

View reviewed changes

Yaminyam requested a review from Copilot December 18, 2024 09:19

Copilot AI reviewed Dec 18, 2024

View reviewed changes

src/ai/backend/manager/models/gql_models/session.py Outdated Show resolved Hide resolved

jopemachine and others added 14 commits December 23, 2024 00:14

feat: Replace sessions, kernels's status_history's type dict …

ed3e4f0

…with `list`

fix: Use SessionId, from_session_id

17277a0

chore: Rename news fragment

dcd9172

chore: fix todo comment

228e5b0

refactor: Use async cmd_main

7634151

fix: Remove useless list copy

e823c6f

fix: Outdated KernelNode.from_row

1699c22

fix: Outdated ComputeSessionNode.from_row

160b323

fix: Missing status_history field

a6354dd

fix: Wrong import

b983e59

fix: Broken scheduled_at field in queryfilter

e2c0139

chore: update GraphQL schema dump

c2209fa

Co-authored-by: octodog <[email protected]>

docs: Rename the news fragment with the PR number\n\n.feature.md -> 3…

b1522d0

…201.feature.md Co-authored-by: octodog <[email protected]>

fix: typos

2b58767

jopemachine force-pushed the topic/12-05-feat_replace_sessions_kernels_s_status_history_s_type_dict_with_list_ branch from 0eb0792 to 2b58767 Compare December 23, 2024 01:00

jopemachine added 2 commits December 23, 2024 02:17

chore: Rename news fragment

dfa5f24

fix: alembic multiple head

8824526

jopemachine requested a review from HyeockJinKim December 23, 2024 04:35

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

feat: Replace `sessions`, `kernels`'s `status_history`'s type `dict` with `list` #3201

feat: Replace `sessions`, `kernels`'s `status_history`'s type `dict` with `list` #3201

jopemachine commented Dec 5, 2024 •

edited by jira bot

Loading

jopemachine commented Dec 5, 2024 •

edited

Loading

jopemachine commented Dec 5, 2024 •

edited

Loading

jopemachine commented Dec 10, 2024

HyeockJinKim Dec 17, 2024

jopemachine Dec 23, 2024

HyeockJinKim Dec 23, 2024

jopemachine Dec 26, 2024

HyeockJinKim Dec 17, 2024

jopemachine Dec 23, 2024 •

edited

Loading

Copilot AI left a comment

	@main.group()
	def session():
	"""Set of compute session operations"""

feat: Replace sessions, kernels's status_history's type dict with list #3201

Are you sure you want to change the base?

feat: Replace sessions, kernels's status_history's type dict with list #3201

Conversation

jopemachine commented Dec 5, 2024 • edited by jira bot Loading

jopemachine commented Dec 5, 2024 • edited Loading

jopemachine commented Dec 5, 2024 • edited Loading

jopemachine commented Dec 10, 2024

HyeockJinKim Dec 17, 2024

Choose a reason for hiding this comment

jopemachine Dec 23, 2024

Choose a reason for hiding this comment

HyeockJinKim Dec 23, 2024

Choose a reason for hiding this comment

jopemachine Dec 26, 2024

Choose a reason for hiding this comment

HyeockJinKim Dec 17, 2024

Choose a reason for hiding this comment

jopemachine Dec 23, 2024 • edited Loading

Choose a reason for hiding this comment

Copilot AI left a comment

Choose a reason for hiding this comment

feat: Replace `sessions`, `kernels`'s `status_history`'s type `dict` with `list` #3201

feat: Replace `sessions`, `kernels`'s `status_history`'s type `dict` with `list` #3201

jopemachine commented Dec 5, 2024 •

edited by jira bot

Loading

jopemachine commented Dec 5, 2024 •

edited

Loading

jopemachine commented Dec 5, 2024 •

edited

Loading

jopemachine Dec 23, 2024 •

edited

Loading