glusterfs.git/xlators/mgmt/glusterd/src/glusterd-rpc-ops.c, branch v3.4.1qa2

glusterd: big lock - a coarse-grained locking to prevent races

2013-04-17T12:48:50+00:00

There are primarily three lists that are part of glusterd process,
that are concurrently accessed. Namely, priv->volumes, priv->peers
and volinfo->bricks_list.

Big-lock approach
-----------------
WHAT IS IT?
Big lock is a coarse-grained lock which protects all three
lists, mentioned above, from racy access.

HOW DOES IT WORK?
At any given point in time, glusterd's thread(s) are in execution
_iff_ there is a preceding, inbound network event. Of course, the
sigwaiter thread and timer thread are exceptions.
A network event is an external trigger to glusterd, via the epoll
thread, in the form of POLLIN and POLLERR.
As long as we take the big-lock at all such entry points and yield
it when we are done, we are guaranteed that all the network events,
accessing the global lists, are serialised.

This amounts to holding the big lock at
- all the handlers of all the actors in glusterd. (POLLIN)
- all the cbks in glusterd. (POLLIN)
- rpc_notify (DISCONNECT event), if we access/modify
  one of the three lists. (POLLERR)

In the case of synctask'ized volume operations, we must remember that,
if we held the big lock for the entire duration of the handler,
we may block other non-synctask rpc actors from executing.
For eg, volume-start would block in PMAP SIGNIN, if done incorrectly.
To prevent this, we need to yield the big lock, when we yield the
synctask, and reacquire on waking up of the synctask.

BUG: 948686
Change-Id: I429832f1fed67bcac0813403d58346558a403ce9
Signed-off-by: Krishnan Parthasarathi 
Reviewed-on: http://review.gluster.org/4835
Tested-by: Gluster Build System 
Reviewed-by: Vijay Bellur

glusterd: Moved node rsp functions to glusterd-utils.c

2013-02-03T19:52:58+00:00

Change-Id: Ib4c4794563a5a694fab16f17c642f788399462f6
BUG: 852147
Signed-off-by: Krishnan Parthasarathi 
Reviewed-on: http://review.gluster.org/4295
Tested-by: Gluster Build System 
Reviewed-by: Kaushal M 
Reviewed-by: Anand Avati

glusterd: log changes in volume stop (and in op sm codepath)

2013-01-18T21:24:17+00:00

This patch makes log changes mostly in the op state machine as also
in volume stop codepath of glusterd.

Changes made:

* Moved log level from INFO to DEBUG, of log messages on the various
  state transitions within a transaction.
  For example, messages of the following kind:
  a. "Sent op req to  peers"
  b. "Received LOCK from uuid: ", etc.

* Changed some of the log messages to give as much information as
  available in case of failure.

* Added logs to identify on which machine lock/stage/commit failed.

* Quite a few s/THIS/this changes.

Also, with this change, log changes in all other volume ops
should (hopefully) boil down to modifying the respective logs in
handler, stage and commit (and brick ops in some cases).

Change-Id: I2b8443042b07fb41a1d12033741f7e156aa6b3da
BUG: 812356
Signed-off-by: Krutika Dhananjay 
Reviewed-on: http://review.gluster.org/4382
Tested-by: Gluster Build System 
Reviewed-by: Kaushal M 
Reviewed-by: Anand Avati

glusterd: log enhancements for volume start

2013-01-04T15:26:25+00:00

* changed some of the log messages to give as much information as
  available in case of failure

* added logs to identify on which machine lock/stage/commit failed

* added macros to represent error strings to maintain uniformity
  among error messages for a given kind of error

* moved error logs wherever possible, from caller to callee to avoid
  code duplication

Change-Id: I0e98d5d3ba086c99240f2fbd642451f175f51942
BUG: 812356
Signed-off-by: Krutika Dhananjay 
Reviewed-on: http://review.gluster.org/4353
Reviewed-by: Pranith Kumar Karampuri 
Reviewed-by: Vijay Bellur 
Tested-by: Vijay Bellur

Revert "glusterd: add "volume label" command"

2012-12-05T00:27:48+00:00

This reverts commit dad16a51ba7e6b1c57529423c57257dbce97ee93

Test script causing "silent" failures during execution.

Change-Id: I26dbb8ed22256071cb415cc3aff572ef8372600e
Reviewed-on: http://review.gluster.org/4268
Tested-by: Gluster Build System 
Reviewed-by: Anand Avati

glusterd: add "volume label" command

2012-12-04T22:42:30+00:00

This command is necessary when the local disk/filesystem containing a brick
is unexpectedly lost and then recreated.  Since 961bc80c, trying to start
the brick will fail because the trusted.glusterfs.volume-id xattr is
missing, and if we can't start it then we can't replace-brick or self-heal
so we're stuck in a permanently degraded state.  This command provides a
way to label the empty brick with the proper volume ID so that further
repair actions become possible.

Change-Id: I1c1e5273a018b7a6b8d0852daf111ddc3fddfdc2
BUG: 860297
Signed-off-by: Jeff Darcy 
Reviewed-on: http://review.gluster.org/4259
Tested-by: Gluster Build System 
Reviewed-by: Anand Avati

BD Backend: CLI commands to create/delete image

2012-11-29T17:39:53+00:00

Cli commands  added to create/delete a LV device.

The following command creates lv in a given vg.
$ gluster bd create :/ 

The following command deletes lv in a given vg.
$ gluster bd delete :/

BUG: 805138
Change-Id: Ie4e100eca14e2ee32cf2bb4dd064b17230d673bf
Signed-off-by: M. Mohan Kumar 
Reviewed-on: http://review.gluster.org/3718
Tested-by: Gluster Build System 
Reviewed-by: Vijay Bellur

glusterd: Moved peer rsp handling functions to glusterd-utils

2012-10-11T18:58:33+00:00

- Moved inner functions used in conjunction with synctask, 'out'.

Change-Id: I7fbfd9881ea58645c4295a9fa7163ddd15a45d2f
BUG: 862834
Signed-off-by: Krishnan Parthasarathi 
Reviewed-on: http://review.gluster.org/4066
Tested-by: Gluster Build System 
Reviewed-by: Anand Avati

glusterd: Fix to log command status at the appropriate time

2012-09-20T15:26:22+00:00

PROBLEM:
In the existing implementation, the success/failure of
execution of a command is decided (and logged) in glusterd
handler functions. Strictly speaking, the logging mechanism
must take into account what course the command takes within
the state machine before concluding whether it succeeded or
failed.

FIX:
This patch attempts to fix the above issue for vol commands.
The format of the log message is as follows:
    for failure:
     : FAILED : 
    for success:
     : SUCCESS

APPROACH (in a nutshell):

 * The command string is packed into dict at cli and sent to
   glusterd.
 * glusterd logs the command status just before doing a
   "submit_reply", which is called (either directly or
   indirectly via a call to glusterd_op_cli_send_response)
   at 2 places for every vol command:
   i. in handler functions, and
   ii. in glusterd_op_txn_complete

In short, the failure of a command in the handler implies the
command has indeed failed. However, its success in the handler
does NOT necessarily mean the command succeeded/will succeed.

Change-Id: I5a8a2ddc318ef2dc2a9699f704a6bcd2f0ab0277
BUG: 823081
Signed-off-by: Krutika Dhananjay 
Reviewed-on: http://review.gluster.org/3948
Tested-by: Gluster Build System 
Reviewed-by: Anand Avati

glusterd: Fix remote processes not being shown in "volume status"

2012-09-18T04:15:46+00:00

Also, refactor function name so it makes more sense.

Change-Id: I3bd3e1bdcacb522ddf4f39d09f73be1c53915362
BUG: 856480
Signed-off-by: Kaushal M 
Reviewed-on: http://review.gluster.org/3933
Reviewed-by: Krishnan Parthasarathi 
Tested-by: Gluster Build System 
Reviewed-by: Amar Tumballi