Skip to content

Meeting minutes of Kimchi technical interlock

sming56 edited this page Jan 7, 2014 · 22 revisions

Date: 2014-01-07

Attendee: Lv Wenruo, Feng Shaohe, Huo Yuxin(On Line), Wang Hongliang, Ding Xing, Shu Ming(On line) Venue: @51352 Recoder: Shu Ming

I) Introduction of the developer UI removal (Shu Ming) II) Kimchi 1.2 Sprint 2 work item discussion, Authentication with privileged and non-privileged user, network detach and attach, CDROM insert and eject III) Round table

AI-1) Investigate on if we can remove the HTML request from the front end. (Shuming) AI-2) Post the network detach and attach interfaces RFC, "/vms?networks=" or "/networks/virbr0/interface". (Feng Shaohe) AI-3) CDROM insert and eject UI implementation (Wang hongliang)

Date: 2013-12-18

Attendee: Huo Yuxin, Wang Hongliang, Shu Ming, Feng Shaohe, Lv Wenruo, Wu Dexin, Zhou Meina Venue: 51358RM@Ring building Recorder: Shu Ming

Kimchi 1.2 feature list to be added and discussion of this list

  1. Storage server target probe(iSCSI & nfs)
  2. Detecting nfs permssion exposed by nfs server
  3. VM cdrom iso ejecting and inserting
  4. boot order customization for VM
  5. network management including detach and attach
  6. template network update
  7. attach to a VM and detach storage volume from a VM
  8. network update
  9. authorization step 2
  10. VM screen shot customization in UI
  11. Federation support
  12. Exception refactors and Kimchi re-layout of the model class.
  13. History host data view

AI-1) Put VM screen shoot customization into github wiki backlog (Shu Ming)

Date: 2013-12-04

Attendee: Huo Yuxin, Wang Hongliang, Shu Ming, Feng ShaoheShao, Lv Wenruo Venue: 51358RM@Ring building Recorder: Shu Ming

  1. Network management API *) Network customization for the VM template. The VM template can be customized with a specific network established. We can list all the available network in the process of VM template creation. Also, a new network can be created in the creating process of the VM template *) Network attachment and detachment for the VM. All the VMs created from one template will attached to the same network and the user may want to move the vm to another network

  2. Two usage scenario VM snapshot *) Host management shutdown and restart need to keep VM state and after the host come back to live, all the VMs in the host should be restored to the state saved *) Snapshoting the disk state and VM can be rollback to the former state. By this way, the VM can abandon all the intermediate changes between.

  3. Security control *) LDAP and AD only save the user ID and password information. All of the permission control information should be stored in Kimchi *) Single sign on support on Kimchi

Date: 2013-11-27

Attendee: Huo Yuxin, Wang Hongliang, Zhou Meina, Shu Ming Venue: 51358RM@Ring building Recorder: Shu Ming

  1. Introduction the history data view proposal by Shu Ming and Yuxin added comments that ISO got standard of time range format

  2. Hongliang was concerned about the new two tasks added to Sprint 2, Software update (host) and Add new yum repository (from ISO and from URL) (host). Need more comments to make the schedule for
    them

  3. Power plugin and power host management for Kimchi. Do we have a clear scope what is power platform specific host management?

  4. Yuxing asked more code review for network patch and Shu ming will take the acttion

AI-1) Shu Ming will take review and test network patch

Date: 2013-11-20

Attendee: Huo Yuxin, Wang Hongliang, Zhou Meina, Ding Xing, Shu Ming, Lv Wenruo, Wu Dexin, Feng Shaohe Venue: 51358RM@Ring building Recorder: Shu Ming

  1. Chinese translation of "isolated network" and "NAT" were discussed. Yuxin

    • "NAT" should not be translated, "isolated work" should be "隔离网络" or "私有网络".
  2. "GET /debugreports" should be authenticated as the same as the other collections like "GET /vms" &etc. Ming and Hongliang

  3. Storage servers probe RFC was discussed

    3-1) Is it necessary to have "POST /storageservers"?

    • POST method can be helpful to save the server name typed by the user and then list the server name automatically when the user try to reuse the server to add another storagepool Dexin and Ming
    • GET /storageservers will return all of the storage servers typed and used by the user. The client will feel confusing for the returned storage server list, if it is all availables to the client?. Yuxin

    3-2) GET /storageservers/host

    • In the backend, the result of GET method should be refreshed after previous GET method's time passed 30 seconds Wenruo
    • What does "GET /storageservers" mean to the user? Does the URI list all the used servers? Should a filter be added to "GET /storageservers"? Yuxin,Dexin, Wenruo
    • URI /storageservers/host/nfs vs URI /storageservers/nfs/host Ming and Wenruo

AI-1: Wenruo will send out the summary of 3) to mailing list

Date: 11-06-2013

Attendee: Shaohe Feng, Hongliang Wang, Meina Zhou, Wenruo Lv(Royce), Huo Yuxin, Shu Ming Recorder: Shu Ming Venue: 51358@Ring building

I) Reviewed the Debugreports REST API, GET and POST method, downloading the report files Two implementation on the interfaces to download the report files were discussed.

  1. The sub-resource of the debug report
  2. A static directory to download the report files. From the perspective of future authorization, the sub-resource is better to control the client's permission to download the report files based on resources. The static directory way rely on the Cherrpy to restrict the report file downloading.

II) Reviewed the XML way to configure the tab pages based on roles. Meina proposed a configuration file to associate the roles with tab pages. Also, Shu Ming pointed out that REST API will also have role based restriction. Hongliang proposed to use an additional configuration file to link the tab pages, roles and REST APIs

III) Host network configuration patch series need review

AI-1) Hongliang started to work on the UI for Debugreports AI-2) Wenruo will review the Host network configuration patch series

Date: 2013 10.29

Meeting room: 51358@Ring building

Attendee: Huo Yuxin, Wang Hongliang, Zhou Meina, Feng Shaohe, Lv Wenruo, Shu Ming

Writter: Wang Hongliang

I) Debug Report Discussed whether to add "generate time". Agreed on adding it.

II) System Basic Information Sheldon will send out patches for system basic information like: processors, memory, etc.

III) API Document Structure Discussed whether the API document API.md is clear enough for everyone including new comers.

IV) Authorization Logged-in user can do any privileged operations.

V) 3rd-party Library Directory Makefile It's time-consuming when developers want to add new JS files. 3rd-party libraries are self-organized and there is no need to set up makefile for them. Suggested setup Makefile files in top directories.

VI) VM Name Edit Currently all resources are managed by VM names. If a VM name is changed, all related resource need upate. Discussed whether VM uuid needs be involved in back-end to avoid it. It will

VII) Dynamically Tabs Loading It will improve extensibility. Tabs will be added/removed dynamically and can be configured. Discussed: 1) how to solve i18n; 2) how to add federation hosts. It has potential to allow user defined UI files (HTML, CSS, JS) dynamically to make it highly extensible.

AI-1) Sheldon will send patches for system basic information.

AI-2) Hongliang will send patches for host tab including debug report(based-on Shu Ming's patch) and system basic information(based-on AI-1).

AI-3) Hongliang will send patches for VM name edit based-on Royce's patch.

AI-4) Meina discuss in further about tabs dynamically loading issue.

Date: 2013 10.23

Meeting room: 5138@Ring building

Attendee: Huo Yuxin, Wang Hongliang, Zhou Meina, Feng Shaohe, Lv Wenruo, Shu Ming

Writter: Shu Ming

I) Deep ISO scanning issues discussed. 1) the most conservative method to not allow Kimchi to use the iso image without permission. After authorization is added, we can add more aggressive way to fix the permission. 2) How to merge the deep scanning result of different users for their directory, the directories may have parent child relationship 3) nfs scanning hanging issue 4) User logout and login to initiate another scanning process that will lead bunches of scanning thread in the backend 5) Scanning task should be able to be canceled and re-started, in future plan.

II) Host statics data 1) Multiple core information in future plan 2) A dedicate thread should be created to collect the host statics

III) IE8 issues Claiming a known issue not supporting IE8

IV) Network interface not discussed

V) jquery UI Makefile issue to be investigated about when we should add distro dependency for the UI related files.

AI-1) Selinux, libvirt and delete storage volume

AI-2) Wen Ruo lead the deep scanning issue discussion

AI-3) Shaohe and Yuxin will look into the Makefile issue

KIMCHI WEEKLY

Meeting Date: Tuesday, September 10th,2013 at 02:00 AM Meeting Room: Room 31355 Ring Building Member attend: Ming Shu, Xin Ding, Shaohe Feng, Wenruo Lv, Hongliang Wang,Meina Zhou Recorder: Meina Zhou

Meeting minutes

1 The design of the authentication UI. (which is good design, and which is unreasonable) 2 The new feature design of the storage. (mainly: create , delete storage pool)

Conclusion 1 Log into kimchi there is a error message “List guest failed” , this issue should be fixed.(need fix by:hongliang mingshu) 2 Log into kimchi after authentication. Redirect to 404 error page.(need fix by: hongliang, shuming) 3 Kimchi should keep the original page when session time out. User should do the action again to complete it.(need fix by: hongliang) 4 List volumes and refresh the storage need to merges together (Wenruo will take this.)

Questions remains: 1 The guest page will refresh every 5s, so it will not time out. The other page will time out, so it will be difference between the tab page. Advise1: Drop the session out. (Need to investigate how important the session time to users.) Advise2: Make each page refresh itself. Advise3: Make the session time out a very big number. (Need more investigation).

2 The storage pool will list all the files in the directory, that is because kimchi don't know the file type in the directory, how to we avoid this issue. (Draft a RFD in the mail list by Meina Zhou).

3 What is the user purpose of delete a storage pool? (Draft a RFD in the mail list by Meina Zhou ) (1)Delete the directory and files in it. (Which is called delete by libvirt.) (2)Delete the directory in config files. (which is called undefined by libvirt.) Advise1: Delete the storage pool in config files when the storage pool do not contain the volumes create by kimchi. Advise2: When the storage pool is real empty, it can be delete from the disk. 4 If we keep the original method of delete storage pool, do we need to change the name from “delete” to “remove”, because it is more reasonable.