glusterfs.git/xlators/performance, branch v3.4.0alpha2

performance/write-behind: guarantee non-overlapping concurrent writes

2013-03-07T07:30:14+00:00

Maintain a list of writes (either written behind or SYNC) which are
currently "in progress" (i.e, STACK_WIND'ed towards server) and hold
off any new STACK_WIND of write (either written behind or SYNC) which
overlaps with any of the "in progress" writes.

This is a guarantee which AFR's eager-lock depends upon (though not
strictly a write-behind requirement)

Change-Id: Icedd0b51b440366a906dc9223d62b7fd6ef2ca03
BUG: 857673
Original-author: Anand Avati 
Signed-off-by: Anand Avati 
Signed-off-by: Jeff Darcy 
Reviewed-on: http://review.gluster.org/4642
Tested-by: Gluster Build System

performance/write-behind: mark fd bad if any written behind writes fail

2013-03-07T05:18:00+00:00

BUG: 765473
Change-Id: I1ddd6ef9f5361aed96f97aa1344823836c6ddecb
Signed-off-by: Raghavendra G 
Reviewed-on: http://review.gluster.org/4630
Tested-by: Gluster Build System 
Reviewed-by: Jeff Darcy

Use proper libtool option -avoid-version instead of bogus -avoidversion

2013-02-07T23:12:56+00:00

Change-Id: I1c9541058c7d07786539a3266ca125a6a15287d8
BUG: 859835
Signed-off-by: Anand Avati 
Original-author: Kacper Kowalik (Xarthisius) 
Signed-off-by: Kacper Kowalik (Xarthisius) 
Reviewed-on: http://review.gluster.org/3967
Tested-by: Gluster Build System

open-behind: translator to perform open calls in background

2013-02-06T21:16:08+00:00

This is functionality peeled out of quick-read into a separate
translator.

Fops which modify the file (where it is required to perform the
operation on the true fd) will trigger and wait for the backend
open to succeed and use that fd.

Fops like fstat() readv() etc. will use anonymous FD (configurable)
when original fd is unopened at the backend.

Change-Id: Id9847fdbfdc82c1c8e956339156b6572539c1876
BUG: 846240
Signed-off-by: Anand Avati 
Reviewed-on: http://review.gluster.org/4406
Tested-by: Gluster Build System 
Reviewed-by: Raghavendra Bhat 
Reviewed-by: Jeff Darcy

performance/write-behind: do not try to take LOCK in forget

2013-02-03T22:09:49+00:00

LOCK attempt in wb_forget is unnecessary

Change-Id: Ibdedc23d0c829c34aedd6fc5bc0e0a584b832514
BUG: 903566
Signed-off-by: Raghavendra Bhat 
Reviewed-on: http://review.gluster.org/4423
Reviewed-by: Anand Avati 
Tested-by: Gluster Build System

quick-read: various fixes

2013-01-29T19:56:42+00:00

- initialize xdata in qr_lookup even if it was NULL from top. This
  allows qr to do its job even if lookup originated from fuse-resolve.c

- extend test cases to include 1 second delay and retry

- fix bug while checking condition for cached unwind

  qr_readv_cached() unwinds if op_ret > 0. Therefore qr_readv()
  must wind to subvol only if !(op_ret > 0) (i.e, op_ret <= 0).

- qr_readv_cached() is using uninitialized @conf pointer. Thanks
  to Raghavendra Bhat for catching this!

Change-Id: Ifaf2ea2685e452210ef9ba3c2d1f2ab51900650c
BUG: 846240
Signed-off-by: Anand Avati 
Reviewed-on: http://review.gluster.org/4452
Reviewed-by: Raghavendra Bhat 
Tested-by: Gluster Build System

performance/io-cache: propagate errors while unwinding frame in

2013-01-29T19:01:34+00:00

                      read path.

Change-Id: Ieb5d592a987e8681d5ec019da309f75e3b207580
BUG: 858242
Signed-off-by: Raghavendra G 
Reviewed-on: http://review.gluster.org/4204
Tested-by: Gluster Build System 
Reviewed-by: Anand Avati

performance/md-cache: add force-readdirp flag to make readdirp configurable

2013-01-29T01:26:37+00:00

md-cache currently transforms all readdir fops into readdirp fops.
This patch creates the 'force-readdirp' configuration flag to
provide control over this behavior. force-readdirp is enabled by
default to maintain current default behavior.

BUG: 903175
Change-Id: Idd70926dec7c271204bdfb11fb052e56d0a39420
Signed-off-by: Brian Foster 
Reviewed-on: http://review.gluster.org/4440
Tested-by: Gluster Build System 
Reviewed-by: Anand Avati

quick-read: refactor

2013-01-28T20:35:57+00:00

- peel out 'open behind' functionality into a separate translator
- issue where, if file size had grown by revalidate, data was not flushed
- removed unnecessary acquistion of table->lock (e.g in qr_lookup())
- keep inode ctx persistent, prune only data (effectively changing the
  order of lock acquisition from INODE -> TABLE)
- validation with readdirplus
- use variable size iobufs to simply cached reads

Change-Id: If1586d0298fd1697ddff9fd7008efb3d286d436a
BUG: 846240
Signed-off-by: Anand Avati 
Reviewed-on: http://review.gluster.org/4403
Tested-by: Gluster Build System 
Reviewed-by: Jeff Darcy

"gcc -pedantic": made 'inline' functions as 'static inline' functions

2013-01-23T19:12:15+00:00

for passing the build with -pedantic flag

Change-Id: I80fd9528321e4c6ea5bec32bf5cdc54cc4e4f65e
BUG: 875913
Signed-off-by: Amar Tumballi 
Reviewed-on: http://review.gluster.org/4186
Reviewed-by: Jeff Darcy 
Tested-by: Gluster Build System 
Reviewed-by: Anand Avati