[dashboard] Remove DataSource.agents with internal kv get. #50025
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Signed-off-by: Ruiyang Wang <[email protected]>
Signed-off-by: Ruiyang Wang <[email protected]>
Signed-off-by: Ruiyang Wang <[email protected]>
jjyao
reviewed
Jan 24, 2025
| @@ -191,76 +198,135 @@ async def _pick_random_agent(self) -> Optional[JobAgentSubmissionClient]: | |||
| """ | |||
| # NOTE: Following call will block until there's at least 1 agent info | |||
| # being populated from GCS | |||
| agent_infos = await self._fetch_agent_infos() | |||
| agent_infos = await self._fetch_all_agent_infos() | |||
There was a problem hiding this comment.
This might be expensive for a large cluster. I feel we can let dashboard gcs client subscribes to nodes table so we have a local cache of nodes.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Many Heads monitor all nodes updates and keep a grpc connection to each agent. This is not needed because most of the use cases are only talking to single agent to read logs etc.
Remove DataSource.agents and remove all practice of "connect to all agents". Instead, when a need comes, make point request to internal kv get to get the corresponding agent grpc address and connect to them.
NodeHead used to add agents to DataSource.agents on new node, and remove agents from DataSource on node removal. Now, it no longer work for the node addition, but it deletes related internal kv keys on node removal.
One notable exception is JobHead which wants to randomly choose from all agents. Use internal_kv_keys to list all known agents.