glusterfs.git/xlators/features/marker/utils/syncdaemon/master.py, branch v3.3.0qa4

gsyncd: do some basic sanitization on logs

2011-07-29T12:24:43+00:00

- exceptions raised by us will be logged as single-line error messages
  (full stack strace is shown only at DEBUG loglevel)
- common/well understood exceptions are mapped to "user-parsable" error logs

Change-Id: I75f1fb848483372364b2093878d9cfed576c9739
BUG: 2778
Reviewed-on: http://review.gluster.com/125
Tested-by: Gluster Build System 
Reviewed-by: Anand Avati

geo-rep: compactify logging

2011-07-28T09:57:18+00:00

Change-Id: I19f6cf539289062b39e3c0f73fa88f4ff9488349
BUG: 2778
Reviewed-on: http://review.gluster.com/80
Tested-by: Gluster Build System 
Reviewed-by: Kaushik BV

syncdaemon: some refactor on monitor

2011-05-09T20:27:03+00:00

- detect faulty state early
- keep the feedback fd in gsyncd module

Signed-off-by: Csaba Henk 
Signed-off-by: Anand Avati 

BUG: 2537 (gsync autorestart)
URL: http://bugs.gluster.com/cgi-bin/bugzilla3/show_bug.cgi?id=2537

syncdaemon: have the monitor kill the worker if it does not connect in 60 sec

2011-04-22T11:05:20+00:00

Signed-off-by: Csaba Henk 
Signed-off-by: Anand Avati 

BUG: 2736 (gsyncd hangs if crash occurs in the non-main thread)
URL: http://bugs.gluster.com/cgi-bin/bugzilla3/show_bug.cgi?id=2736

syncdaemon: force termination for unhandled exception in any thread

2011-04-13T11:43:06+00:00

Signed-off-by: Csaba Henk 
Signed-off-by: Anand Avati 

BUG: 2736 (gsyncd hangs if crash occurs in the non-main thread)
URL: http://bugs.gluster.com/cgi-bin/bugzilla3/show_bug.cgi?id=2736

syncdaemon: streamline user-visible messages (and control flow a bit, too)

2011-04-13T11:43:02+00:00

Signed-off-by: Csaba Henk 
Signed-off-by: Anand Avati 

BUG: 2644 (no volume-mark, if the behaviour persists have to check if master gsyncd is running.)
URL: http://bugs.gluster.com/cgi-bin/bugzilla3/show_bug.cgi?id=2644

syncdaemon: minor fixes

2011-04-13T11:42:44+00:00

- whitespace fix
- remove superfluous imports

Signed-off-by: Csaba Henk 
Signed-off-by: Anand Avati 

BUG: 1570 (geosync related changes)
URL: http://bugs.gluster.com/cgi-bin/bugzilla3/show_bug.cgi?id=1570

syncdaemon: fix setting None as new master

2011-04-11T10:52:25+00:00

We checked for the event of a new master being adopted by:
- checking for state change in list of masters
- checking if our chosen master is a foreign one
and if these held, then wrote out the actually available master to config

However this was flaky: if
(1) native master changed
(2) there was already a foreign master adopted
(3) actually the master was not available
then we wrote out the None to config.

This could occur if:
- on syncd startup, native master just changes from None to the volume native
  [implies (1)]
- foreign master adopted from earlier run via config [implies (2)]
- actually foreign master was not available [implies (3)]

Signed-off-by: Csaba Henk 
Signed-off-by: Anand Avati 

BUG: 2670 (strange behaviour with intermediate-master and slave)
URL: http://bugs.gluster.com/cgi-bin/bugzilla3/show_bug.cgi?id=2670

syncdaemon: add monitor mode to support autorestart

2011-04-04T15:02:27+00:00

Signed-off-by: Csaba Henk 
Signed-off-by: Vijay Bellur 

BUG: 2537 (gsync autorestart)
URL: http://bugs.gluster.com/cgi-bin/bugzilla3/show_bug.cgi?id=2537

syncdaemon: give some refactoring to cascading code

2011-04-04T15:02:18+00:00

- expiry check of foreign volinfo moved back to GLUSTERServer,
  so that under the hood we can removexattr the expired ones;
  a nice side-effect is that we can use the same dict layout
  for foreign and native volinfo (ie., foreign needs no
  timeout field)
- get_volinfo() is renamed to get_sys_volinfo() and most of the logic
  is stripped off of it (what remained there is the check
  against foreign master ambiguity)
- volinfo transition logic is cut out to an almost purely functional
  static method (only impurity is the exeption raised upon
  forbidden volinfo change)
- ping renamed to keep-alive, as something called "ping" is
  not supposed to have payload (yeah, keep-alive is a bit fishy
  on this front too, but could not come up with better...)

Signed-off-by: Csaba Henk 
Signed-off-by: Vijay Bellur 

BUG: 2535 (gsync cascading)
URL: http://bugs.gluster.com/cgi-bin/bugzilla3/show_bug.cgi?id=2535