Re: [pnfs] Linux pNFS status meeting 05/13

Re: [pnfs] Linux pNFS status meeting 05/13

[J. Bruce Fields]

On Wed, May 12, 2010 at 12:35:18PM -0700, Marc Eshel wrote:
> Meeting on Thursday 05/13/10 at 9:30 AM pacific time (12:30 PM UMICH time)

Rough agenda:

1. upstream status/merge plans (trond, bfields)
2. State of pNFS tree (bhalevy)
3. Mailing list change
4. Client
	4.1. Client pNFS status
	4.2. Client state management design documentation (trond)
5. Server
	5.1. Todo's for minimal 4.1 server (bfields)
		- Benny's patches
		- pynfs testing
	5.2. pnfs
6. redhat status (steved)
7. bakeathon plans
8. new business
9. next meeting

--b.

Re: [pnfs] Linux pNFS status meeting 05/13

[J. Bruce Fields]

On Wed, May 12, 2010 at 04:08:01PM -0400, bfields wrote:
> On Wed, May 12, 2010 at 12:35:18PM -0700, Marc Eshel wrote:
> > Meeting on Thursday 05/13/10 at 9:30 AM pacific time (12:30 PM UMICH time)
> 
> Rough agenda:

Apologies, I've been busier than usual and forgot to send out my notes:

1. upstream status/merge plans (trond, bfields)
	- rc7 released
	- bfields: reclaim complete, benny's patches
	- Trond: cleaning up branches, merged 2.6.35 stuff in (stack
	  reduction, Chuck's rtc-resolution patches).  Haven't merged in
	  access & auth cache -shrinker changes.  Also: state-recovery
	  thread deadlock still being worked.  Rearranged rpc-task
	  structure fields to shrink struct size.  That done, will send
	  changelog to list.

2. State of pNFS tree (bhalevy)
	- [distracted, missed some notes]
	- note also cleaned up gfs2 branch

3. Mailing list change
	- no problems reported, except Benny getting repeatedly bounced
	  off?  He'll investigate farther.  No response from postmaster.

4. Client
	4.1. Client pNFS status
		- stable/unstable write-discussion.
		- Andy notes: usually tests with writes to gfs2 turned
		  on, even though not really optimal (and will stay turned
		  off in mainstream).
		- Benny: working on some io path bugs; will discuss on
		  mailing list bugs.  Ricardo: note Fred may be working
		  on same area.  (E.g.: on "done" path, two closes (one
		  of layout_commit context, one layout return on close
		  path?) were racing with each other?  Looks like there
		  may be patches that fix that.  Also saw stable
		  object-layout write problems.)
		- Alexandros doing some work on layout-management, close
		  to sending.
		- what about ipv6?: reports are that it's work; not sure
		  anyone's been thinking about it hard, though.
		  (What about device cache?  Andy should take a look.
		  What about gfs2 data server references?  Ricardo says
		  this was all "priority B" on client.)
	4.2. Client state management design documentation (trond)
		- some progress last week, none this.
5. Server
	5.1. Todo's for minimal 4.1 server (bfields)
		- Benny's patches
		- pynfs testing
	5.2. pnfs
		- question about exports.
		- exofs; did some testing, no report yet (distracted by
		  some bugs)

6. redhat status (steved)
	- nothing to report
7. bakeathon plans
	- ann arbor: going fine
	- boston: still waiting on hotel details.
8. new business
	- none
9. next meeting
	- same time (note Benny and Boaz expect to miss it).

--b.