glusterfs.git/xlators/mgmt/glusterd/src/glusterd-brick-ops.c, branch v3.3.0qa20

glusterd: prevent adding bricks already in use

2012-01-13T05:43:24+00:00

The add-brick command now checks if the brick provided for
add-brick is used in any volumes, even if the volume was
never started by looping through the brick lists of all
volumes.

Change-Id: I15035d41d91386448a3e3d4063d909b880288681
BUG: 771831
Signed-off-by: Rajesh Amaravathi 
Reviewed-on: http://review.gluster.com/2607
Tested-by: Gluster Build System 
Reviewed-by: Krishnan Parthasarathi 
Reviewed-by: Amar Tumballi

glusterd: Fixed add-brick error message.

2012-01-05T08:52:19+00:00

Change-Id: I9260d5a0203ae2990d37b30bcc7c94c81c1e168e
BUG: 770678
Signed-off-by: Krishnan Parthasarathi 
Reviewed-on: http://review.gluster.com/2581
Tested-by: Gluster Build System 
Reviewed-by: Amar Tumballi

glusterd: Added checks for remove-brick handler.

2012-01-04T19:32:01+00:00

Removing bricks in a plain stripe-replicate is failed
if replica count is not reduced explicitly.

Change-Id: I6d0de4862595744d1d1998b9a287c34c53d7fe5f
BUG: 770561
Signed-off-by: Krishnan Parthasarathi 
Reviewed-on: http://review.gluster.com/2577
Tested-by: Gluster Build System 
Reviewed-by: Jeff Darcy 
Reviewed-by: Vijay Bellur

glusterd: Fixed add-brick handler algorithm.

2012-01-03T04:49:58+00:00

In the fix for 765774, setting the op dictionary with modified stripe
and replica count was (accidentally) removed affecting the correctness.
The stage and op functions of add-brick rely on the fact that stripe/replica
count are zero if there is no change in them during the add-brick operation.

Change-Id: I55a23878e618b355009ffd0fce8c4904dbb70691
BUG: 770914
Signed-off-by: Krishnan Parthasarathi 
Reviewed-on: http://review.gluster.com/2548
Tested-by: Gluster Build System 
Reviewed-by: Vijay Bellur

glusterd: Fixed spurious checks in add_brick.

2011-12-20T05:42:30+00:00

Previous algo,
 - didn't allow volume type change from plain distribute
   to distribute-[stripe][-][replicate].
 - performed 'add-brick' on temporary volinfo without correct
   initialisations. This was superflous.

Change-Id: If5d5d78e59bd04685df9c27974f715df5f0b43f3
BUG: 765774
Signed-off-by: Krishnan Parthasarathi 
Reviewed-on: http://review.gluster.com/804
Tested-by: Gluster Build System 
Reviewed-by: Amar Tumballi

XDR: cli-glusterd xdr consolidation

2011-11-16T09:42:29+00:00

By using only 1 xdr struct for request and 1 xdr struct for response,
we will be able scale better and also be able to parse the o/p better

For request use-
        gf1_cli_req - contains dict
For response use-
        gf1_cli_rsp - conains op_ret, op_errno, op_errstr, dict

Change-Id: I94b034e1d8fa82dfd0cf96e7602d4039bc43fef3
BUG: 3720
Reviewed-on: http://review.gluster.com/662
Tested-by: Gluster Build System 
Reviewed-by: Amar Tumballi

changing of the volume types after volume is created.

2011-10-28T12:42:08+00:00

with this patch, one can do all the below operations:

on a simple distribute volume 'test' with bricks A,B, add bricks C and D:

 bash# gluster volume add-brick test replica 2 C D

now the volume 'test' becomes distribute-replicate, with pairs A-C, B-D.
same holds good for stripe.

and now, one can change the replica count 2 on volume 'test' to 4 like below:

 bash# gluster volume add-brick test replica 4 E F G H

now, volume has 'replica' count 4, with pairs being 'A-C-E-F' and 'B-D-G-H'.

after this, if user wants to reduce the replica count, he can do

 bash# gluster volume remove-brick test replica 3 C H # (one from each pair)

This should work to reduce the replica count

Change-Id: I4f109141a9ba7be4a92a5f5810a4163baeb92973
BUG: 3305
Reviewed-on: http://review.gluster.com/158
Tested-by: Gluster Build System 
Reviewed-by: Vijay Bellur

glusterd: take off 'auto-commit' behavior of decommissioning

2011-10-11T10:37:51+00:00

currently if 'remove-brick  start' is given, after all
the data is migrated, in the _cbk() function from rebalance
complete we used to re-write the volume file without the bricks.
But the flaw of this method is that, currently the _cbk() function
is called on only the node on which rebalance actually happens,
hence other nodes are not aware of this behavior. So, had to remove
the 'auto-commit', instead we want the user to issue a command
'remove-brick  commit' (or 'force') explicitely to remove
the bricks

Change-Id: Ifdac1292095029429da502aa6e4c0cf4ee124990
BUG: 1952
Reviewed-on: http://review.gluster.com/551
Tested-by: Gluster Build System 
Reviewed-by: Shishir Gowda 
Reviewed-by: Vijay Bellur

glusterd: allow non-related directories with same prefix

2011-09-29T16:16:13+00:00

Earlier, using two directories with names like /test and /test1
failed volume creation. This patch sees that the next character is
checked to be '/' to conclude that they are recursive.
Eg: /test and /test2 succeed, while /test and /test/alpha does not.

Change-Id: Idee1e15b7e0a0d0965d1aea910b52f1da286d64b
BUG: 3645
Reviewed-on: http://review.gluster.com/511
Tested-by: Gluster Build System 
Reviewed-by: Amar Tumballi 
Reviewed-by: Krishnan Parthasarathi

glusted: while doing add-brick, handle distribute volume separately

2011-09-27T12:35:03+00:00

so the logic of different type of counts etc won't come into action,
which would fail the add-brick operation.

Change-Id: I5c8769d6d1c2433ffb084e5dce9c1e01678cff40
BUG: 3630
Reviewed-on: http://review.gluster.com/505
Tested-by: Gluster Build System 
Reviewed-by: Krishnan Parthasarathi