Difference between revisions of "2015-06-22 Protocols Document meeting minutes"

From OpenSFS Wiki
Jump to navigation Jump to search
(Created page with "Category:SFS-DEV-005 == Attendees == * Olaf Faaland (LLNL) * Chris Morrone (LLNL) * Sarp Oral (ORNL) * Doug Oucharek (Intel) * Andrew Uselton (Intel) * Peter Jones (Inte...")
 
Line 4: Line 4:
  
 
* Olaf Faaland (LLNL)
 
* Olaf Faaland (LLNL)
 +
* Chris Horn (Cray)
 
* Chris Morrone (LLNL)
 
* Chris Morrone (LLNL)
 +
* Colin Faber (Seagate)
 +
* Justin Miller (Cray)
 +
* Andrew Uselton (Intel)
 +
* Richard Henwood (Intel)
 +
 +
Apologies:
 
* Sarp Oral (ORNL)
 
* Sarp Oral (ORNL)
 
* Doug Oucharek (Intel)
 
* Doug Oucharek (Intel)
* Andrew Uselton (Intel)
 
 
* Peter Jones (Intel)
 
* Peter Jones (Intel)
* Richard Henwood (Intel)
 
 
* Andreas Dilger (Intel)
 
* Andreas Dilger (Intel)
* Colin Faber (Seagate)
 
  
 
== Agenda ==
 
== Agenda ==
  
+ progress since last week.
+
=== progress since last week ===
  + statfs isn't done - need to integrate Chris's feedback on error codes
+
   
and highlight differences between MDS_STATFS and OST_STATFS op codes,
+
* completed milestone review and sent out our plans.
hyperlinked sequence diagram (example attached). After this, it will be
+
* total number of fs items is ~50
done.
 
+ further feedback on the document?
 
+ initial discussion on state machine: our preference is file create
 
 
 
  
 +
Chris highlighted the feedback he had already provided in email: he thought our progress advised that Recovery operations are an order of magnitude more complex.
  
=== Brief status ===
+
AR: Richard to provide a list of the fs semantic operations that we will document.
  
* Patch from Mike Pershin landed.
+
=== getting to done with statfs ===
* Reviews from Doug and Jinshan.
 
  
=== Progress on detailing section ===
+
The complete task list is:
  
This task is about taking a particular RPC an adding in more detail to create a 'gold standard' that can be used as a template in future.
+
# integrate Chris's feedback on error codes
 +
# highlight differences between MDS_STATFS and OST_STATFS op codes
  
+ settled on statfs as the detailed item.
+
Chris shared that from his reading the error codes don't look right. The call agreed that no further work should be performed on the hyperlinked sequence diagram and all efforts concentrated on completing the right level of detail.
+ Andrew understands this task as looking at what happens at the boundary conditions for RPC fields.
 
  
This is what Chris is expecting.
 
  
Andrew is expecting to provide a version of statfs for review in a day or two. For clarification, a further discussion of the error conditions when statting a file took place. The questions that should be considered include:
+
=== initial discussion on state machine: our preference is file create ===
  
* when you try and stat a file, what happens when the file doesn't exist?
+
The call agreed that this was premature while the detail level was still being worked on.
* what kinds of messages get sent when an error happens?
+
=== closing comments ===
  
Sarp request progress milestones need to be shared so OpenSFS can
+
The closing comments from Chris and Colin were:
  
Richard said he would prepare suitable milestones for next weeks meeting.
+
Process change needs to be made and code experts are not participating sufficiently to this project.

Revision as of 12:20, 22 June 2015


Attendees

  • Olaf Faaland (LLNL)
  • Chris Horn (Cray)
  • Chris Morrone (LLNL)
  • Colin Faber (Seagate)
  • Justin Miller (Cray)
  • Andrew Uselton (Intel)
  • Richard Henwood (Intel)

Apologies:

  • Sarp Oral (ORNL)
  • Doug Oucharek (Intel)
  • Peter Jones (Intel)
  • Andreas Dilger (Intel)

Agenda

progress since last week

  • completed milestone review and sent out our plans.
  • total number of fs items is ~50

Chris highlighted the feedback he had already provided in email: he thought our progress advised that Recovery operations are an order of magnitude more complex.

AR: Richard to provide a list of the fs semantic operations that we will document.

getting to done with statfs

The complete task list is:

  1. integrate Chris's feedback on error codes
  2. highlight differences between MDS_STATFS and OST_STATFS op codes

Chris shared that from his reading the error codes don't look right. The call agreed that no further work should be performed on the hyperlinked sequence diagram and all efforts concentrated on completing the right level of detail.


initial discussion on state machine: our preference is file create

The call agreed that this was premature while the detail level was still being worked on.

closing comments

The closing comments from Chris and Colin were:

Process change needs to be made and code experts are not participating sufficiently to this project.