MaverickFinishUpstart

Differences between revisions 4 and 5
Revision 4 as of 2010-10-18 14:22:09
Size: 2620
Editor: 82-69-40-219
Comment: fix spec link
Revision 5 as of 2010-11-15 14:58:38
Size: 5697
Editor: 82-69-40-219
Comment: gobby dump
Deletions are marked like this. Additions are marked like this.
Line 63: Line 63:
{{{
Please document the outcome of this session at:
https://wiki.ubuntu.com/UDSProceedings/N/PackageSelectionAndSystemDefaults#Finish Upstart

Last UDS (recap): Stable version of Upstart in Ubuntu now for a few releases, working out reasonably well, but there are a number of things we need to fix (the fact that mountall is needed, user services, etc.). Used last UDS to get a sense of what complaints were from various people and groups, and made sure that appropriate bugs were filed.

Upstart's design is too simple. The goal of the next version is to fix problems based on deployment experience while retaining the "Upstartishness", and reach an elegant, simple design that we don't need to change again: i.e. 1.0.

https://bugs.edge.launchpad.net/upstart/+bugs
https://bugs.edge.launchpad.net/upstart/+bug/447654

BUG: upstart events can trigger apparent deadlocks

Biggest upstart bug:
  job maurice has "start on A and B"
  then emit event A
  process emitting event A hangs until event B occurs

The issue is when we use or:

  start on A and (B or C)

If you now emit A, then B, then C, the C will block waiting for a _second_ A.

(Diagram: http://bit.ly/ddfeGG)

The solution is to make states from events. And then jobs can wait on states in addition to events.

BUG: upstart keeps files open on /

BUG: pid tracking can be defeated which leads to upstart breakage

we should be using the proc connector to track children, this would resolve the tracking issue

BUG: service which is slow to start can appear started

Jobs which are starting slowly can appear started to a subsequent start of that job which makes that second start return immediately when the service is not actually started. This for example can trigger gdm to start too early.

This can be fixed by queuing new events against a job when a job is transitioning.

ISSUE: its very hard to find out why jobs are running

We want to be able to generate a dependency graph from a boot to find out why jobs have run. A solution here would also allow an interactive boot.

ISSUE: chroots do not work, as you talk to the 'wrong' upstart

likely solution, tell upstart about 'chroots' so that it can track them and use the right job tree
 * if upstart is explicitly told about chroot/etc/init, then automatically start jobs in that chroot at boot
 * otherwise, assume /etc/init relative to /proc/PID/root when event received

Proposed Changes:
 - add the concepts of states which are based on events and persist beyond the event, which jobs can depend on
 - child tracking should use proc connector
 - events should be queued against jobs when the job is transitioning
 - overrides to local configuration of jobs without editing them
 - new hook on starting * to allow tracking of job dependancies
 - upstart will know about chroots and make itself available in there if it is going to start jobs automatically; and use the root directory for local start within a chroot to start the right jobs

Need to get buy in for these features
}}}

Summary

This should provide an overview of the issue/functionality/change proposed here. Focus here on what will actually be DONE, summarising that so that other people don't have to read the whole spec. See also CategorySpec for examples.

Release Note

This section should include a paragraph describing the end-user impact of this change. It is meant to be included in the release notes of the first release in which it is implemented. (Not all of these will actually be included in the release notes, at the release manager's discretion; but writing them is a useful exercise.)

It is mandatory.

Rationale

This should cover the _why_: why is this change being proposed, what justifies it, where we see this justified.

User stories

Assumptions

Design

You can have subsections that better describe specific parts of the issue.

Implementation

This section should describe a plan of action (the "how") to implement the changes discussed. Could include subsections like:

UI Changes

Should cover changes required to the UI, or specific UI that is required to implement this

Code Changes

Code changes should include an overview of what needs to change, and in some cases even the specific details.

Migration

Include:

  • data migration, if any
  • redirects from old URLs to new ones, if any
  • how users will be pointed to the new way of doing things, if necessary.

Test/Demo Plan

It's important that we are able to test new features, and demonstrate them to users. Use this section to describe a short plan that anybody can follow that demonstrates the feature is working. This can then be used during testing, and to show off after release. Please add an entry to http://testcases.qa.ubuntu.com/Coverage/NewFeatures for tracking test coverage.

This need not be added or completed until the specification is nearing beta.

Unresolved issues

This should highlight any issues that should be addressed in further specifications, and not problems with the specification itself; since any specification with problems cannot be approved.

BoF agenda and discussion

Use this section to take notes during the BoF; if you keep it in the approved spec, use it for summarising what was discussed and note any options that were rejected.

Please document the outcome of this session at:
https://wiki.ubuntu.com/UDSProceedings/N/PackageSelectionAndSystemDefaults#Finish Upstart

Last UDS (recap): Stable version of Upstart in Ubuntu now for a few releases, working out reasonably well, but there are a number of things we need to fix (the fact that mountall is needed, user services, etc.).  Used last UDS to get a sense of what complaints were from various people and groups, and made sure that appropriate bugs were filed.

Upstart's design is too simple.  The goal of the next version is to fix problems based on deployment experience while retaining the "Upstartishness", and reach an elegant, simple design that we don't need to change again: i.e. 1.0.

https://bugs.edge.launchpad.net/upstart/+bugs
https://bugs.edge.launchpad.net/upstart/+bug/447654

BUG: upstart events can trigger apparent deadlocks

Biggest upstart bug:
  job maurice has "start on A and B"
  then emit event A
  process emitting event A hangs until event B occurs

The issue is when we use or:

  start on A and (B or C)

If you now emit A, then B, then C, the C will block waiting for a _second_ A.  

(Diagram: http://bit.ly/ddfeGG)

The solution is to make states from events.  And then jobs can wait on states in addition to events.

BUG: upstart keeps files open on /

BUG: pid tracking can be defeated which leads to upstart breakage

we should be using the proc connector to track children, this would resolve the tracking issue

BUG: service which is slow to start can appear started

Jobs which are starting slowly can appear started to a subsequent start of that job which makes that second start return immediately when the service is not actually started. This for example can trigger gdm to start too early.

This can be fixed by queuing new events against a job when a job is transitioning.

ISSUE: its very hard to find out why jobs are running

We want to be able to generate a dependency graph from a boot to find out why jobs have run.  A solution here would also allow an interactive boot.

ISSUE: chroots do not work, as you talk to the 'wrong' upstart

likely solution, tell upstart about 'chroots' so that it can track them and use the right job tree
 * if upstart is explicitly told about chroot/etc/init, then automatically start jobs in that chroot at boot
 * otherwise, assume /etc/init relative to /proc/PID/root when event received

Proposed Changes:
 - add the concepts of states which are based on events and persist beyond the event, which jobs can depend on
 - child tracking should use proc connector
 - events should be queued against jobs when the job is transitioning
 - overrides to local configuration of jobs without editing them
 - new hook on starting * to allow tracking of job dependancies
 - upstart will know about chroots and make itself available in there if it is going to start jobs automatically; and use the root directory for local start within a chroot to start the right jobs

Need to get buy in for these features


CategorySpec

FoundationsTeam/Specs/MaverickFinishUpstart (last edited 2010-12-15 09:22:10 by 92)