You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Consider having a daemonless lima-guestagent send-event for plain mode. The daemonless GA can also inform additional information such as IP address -- @jandubois
For plain mode, ip command should be executed, or lima-guestagent should inspect the IP using Go stdlib? -- @songponssw
The same Go implementation should be used regardless to plain or not -- @jandubois
Seems fine as long as it is opt-in -- @AkihiroSuda
Mid-term plan, not targeted for v2.0 -- @jandubois
The team discussed implementing eBPF for monitoring system cores and improving Kubernetes API performance, with plans to convert an existing PR to a draft and have Balaji work on the eBPF solution. They reviewed the timeline for Lima 2.0, agreeing it was a medium-term plan not ready for the upcoming KubeCon presentation, while also discussing various technical issues including disk-related features and YAML formatting changes. The team concluded by addressing implementation details for IP address retrieval and discussing proposed changes to the "disrupts" option, emphasizing the importance of maintaining backward compatibility.
Next Steps
Anders to rebase the VM Opts PR and notify the team when it's ready for review
Jan to update the URL scheme PR after Ola's PR is merged
Balaji to work on implementing the eBPF approach for monitoring system calls
Akihiro to review the proposal for fixing the --yes option inconsistency
Ansuman to work on implementing LibKrun after current tasks
Songpon to update the PR for getting IP addresses with a consistent implementation
Team to review and prune issues from the Lima 2.0 milestone to meet the KubeCon release timeline
eBPF for Kubernetes API Performance:
The team discussed implementing eBPF for monitoring system cores and stimulating the ticker to improve Kubernetes API performance. They agreed to keep the current Kubernetes API client but adjust the 3-second ticker to use eBPF to stimulate the ticker for C-scores. Jan suggested converting the existing PR to a draft and having Balaji work on implementing the eBPF solution. They also discussed the need to eliminate binary props and reduce CPU-specific code in the current implementation. The team briefly touched on the possibility of creating a mechanism for sending events in provisioning mode, but decided to focus on non-provisioning instances for now.
Lima 2.0 Planning and Prioritization:
Jan and Akihiro discussed the timeline for Lima 2.0, agreeing that it is a medium-term plan not ready for the upcoming KubeCon presentation. They decided to review all issues marked for the 2.0 milestone and prioritize easy wins that can still be included, while moving less critical items to later releases. Akihiro mentioned ongoing work on integrating AI and addressing issues with the MCP SERP command, which they plan to showcase at KubeCon. They also briefly touched on potential changes to the drive cycle for better driver management.
Disk Feature Implementation Discussion:
The team discussed several technical issues, focusing on the implementation of a disk-related feature. Akihiro explained his desire to eliminate the "base disk" concept, which Jan clarified was not a significant concern as it doesn't consume additional disk space on macOS. They agreed to move forward with the implementation for Broker 2, with Ansuman planning to handle it after completing LibKrun. The team also discussed the timeline for releasing Glimmer 2.0 and the potential need to extend the driver API to support LibKrun. Anders suggested that the driver could be released separately from the main project, which Akihiro agreed was a viable option.
YAML Formatting and Template Changes:
The team discussed changes to YAML formatting and template embedding, with Anders explaining that while there were some casing issues with Go structs, these could be addressed by documenting the changes and potentially moving documentation to the wiki. They agreed that for Lima 2.0, they could make incompatible changes if necessary, though they would aim to maintain backwards compatibility where possible. The team also discussed PR reviews, with Anders mentioning he would need to leave in 15 minutes, and Jan noted that a PR for plugin commands was awaiting another PR to be merged before proceeding.
gRPC to SSH Port Transition:
Akihiro and Jan discussed replacing gRPC port forwarding with SSH for TCP mode to improve stability, while keeping gRPC for UDP and iNotify support. They agreed to investigate the implementation details and consult Balaji about iNotify's dependency on gRPC. Akihiro proposed merging recent changes to BFC for SSH acceleration and addressing a static port routing issue. They decided to proceed with merging these changes after further review.
Disrupts Option Proposal Review:
Jan proposed changes to the "disrupts" option to improve clarity and consistency, suggesting a new option for scripting purposes while making the existing "yes" option hidden. Akihiro and Anders agreed to review the proposal, acknowledging potential backward compatibility issues. Jan emphasized the importance of maintaining backward compatibility and suggested deprecating the current "yes" option while providing alternative functionality. The team agreed to further discuss the changes and consider Jan's detailed suggestions.
IP Implementation Standardization Discussion:
The team discussed implementation details for IP address retrieval, with Songpon seeking clarification on using identical implementations across different modes. Jan emphasized the importance of using identical implementations whenever possible, and Songpon agreed to update his PR accordingly. The conversation ended with a reminder about the next meeting on October 16th, which will focus on approvals for Lima browser to engineering.
They reviewed the timeline for Lima 2.0, agreeing it was a medium-term plan not ready for the upcoming KubeCon presentation
Lima 2.0 should be ready before KubeCon.
eBPF for monitoring system cores and stimulating the ticker to improve Kubernetes API performance.
Kubernetes is not covered in the current proposal, and Kubernetes does not suffer from the API performance issue so far.
They agreed to keep the current Kubernetes API client but adjust the 3-second ticker to use eBPF to stimulate the ticker for C-scores.
Not sure what "C-scores" means
MCP SERP
MCP SERVE
Broker 2
Not sure what "Broker 2" means
Glimmer 2.0
Lima 2.0
BFC for SSH
Not sure what "BFC" means. Probably vsocks?
Jan proposed changes to the "disrupts" option to improve clarity and consistency, suggesting a new option for scripting purposes while making the existing "yes" option hidden.
reacted with thumbs up emoji reacted with thumbs down emoji reacted with laugh emoji reacted with hooray emoji reacted with confused emoji reacted with heart emoji reacted with rocket emoji reacted with eyes emoji
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
-
How to join
The meetings are open to anyone, not just for maintainers and contributors.
Attendees
@AkihiroSuda @unsuman @olamilekan000 @jandubois @afbjorklund @dharsanb @songponssw
Agenda
Note taker: @AkihiroSuda
Subscribe to container engine API for published ports #4021
Guest monitoring #4037
lima-guestagent send-event
for plain mode. The daemonless GA can also inform additional information such as IP address -- @janduboisip
command should be executed, orlima-guestagent
should inspect the IP using Go stdlib? -- @songponsswDefine "MCP Sandbox Interface" and implement
limactl mcp serve
#3744Lima 2.0 release status
limactl shell --sync-host-workdir
(prevents AI agents from breaking the host files) #3711--yes
option is inconsistent and confusing #3995--yes
, needs to be decided before v2 GAlimactl shell --preserve-env
#4036Lima v2.0 schedule
(Feel free to suggest more agenda in the comment form)
Tasks
(See above)
Beta Was this translation helpful? Give feedback.
All reactions