module doesn't provide quorum itself, merely a framework for setting and
querying it. I envisage YKD plugging into this rather than straight into
sync() eventually.
I've plugged this into the sync() routines rather than replacing them so
that quorum is itself a VSF, rather than a replacement - I'm not sure if
that is best or not. Opinions are welcome.
I've added an extra enum member to the service_handler so that we can
send IPC messages when the cluster isn't quorate. This will default to
NO (as now) but allows us to query and set quorum when we don't have it
.. a useful feature !
git-svn-id: http://svn.fedorahosted.org/svn/corosync/trunk@1674 fd59a12c-fef9-0310-b244-a6a79926bd2f
Also add a notification callback for reload. Both the start and the stop
are notified so that the user can decide what to do with the flurry of
updates that occur.
Note than the reload callback notification MUST be registered against
OBJECT_PARENT_HANDLE. Registering it lower down the hierarchy makes no sense.
git-svn-id: http://svn.fedorahosted.org/svn/corosync/trunk@1673 fd59a12c-fef9-0310-b244-a6a79926bd2f
This patch causes the flow control state in the library to be set
properly when the flow control is turned off (disabled). Then it can be
read properly by the flow control apis.
This also fixes the case where the application is no longer sending
messages and it has already dispatched all its received messages
before flow control is disabled.
Also, CPG response messages with a TRY_AGAIN error did NOT contain
a valid flow control state value. This meant the library could get
stuck with flow control enabled (flow control was never enabled
for the EXEC, so no disable event occurred).
This case was hit when a new node was joining - sync_in_process()
resulted in a TRY_AGAIN for error cpg_mcast_joined).
Also, in message_handler_req_exec_cpg_mcast() the state passed
back to the library defaulted to disabled for messages received
from another node (even if flow control was still enabled)
- this meant if multiple nodes were sending CPG messages,
then the library flow control state flip-flopped between
enabled and disabled.
Author: Steven Dake <sdake@redhat.com> &
Tim Beale <tim.beale@alliedtelesis.co.nz>
git-svn-id: http://svn.fedorahosted.org/svn/corosync/trunk@1667 fd59a12c-fef9-0310-b244-a6a79926bd2f
In some cases, the syslog() call may block. If this happened, the AIS worker
thread would block waiting on the syslog() call. However, the worker thread
is still holding a logsys mutex, which is needed to enqueue any more log
messages from the other AIS threads. So the main AIS thread blocks waiting
on the logsys mutex. This can then cause aisexec on other nodes to detect
the node as leaving the cluster.
Author: Tim Beale <tim.beale@alliedtelesis.co.nz>
git-svn-id: http://svn.fedorahosted.org/svn/corosync/trunk@1665 fd59a12c-fef9-0310-b244-a6a79926bd2f
confdb subsystems.
This is useful to provide atomic counters (ag handle numbers) for
long-running (though not persistent) connections. It's not currently
possible via confdb to atomically get a new number from objdb due to the
lack of locking. Doing it via increment operations in the IPC thread
provides enough atomicity to make it useful. Fabio has already
identified a use for these calls.
It could also provide some form of basic co-operative locking mechanism
for IPC-using processes (not direct objdb calls).
git-svn-id: http://svn.fedorahosted.org/svn/corosync/trunk@1662 fd59a12c-fef9-0310-b244-a6a79926bd2f
the confdb library to use the new find_create/find_next/find_destroy API
calls instead.
I've kept the libcondfb API the same as before with the single change of
adding a confdb_object_find_destroy to tidy up the find handle after
use. If you don't call this then libcondfb will do it for you when
confdb_finalize is called.
git-svn-id: http://svn.fedorahosted.org/svn/corosync/trunk@1660 fd59a12c-fef9-0310-b244-a6a79926bd2f
When making my_proc_list smaller, ensure that the
now non-used entries are zero-ed out. There are some suspect
assert's that assume that there is always 2 entries in the list.
These fail when my_proc_list is reduced to 1 entry (and the
valid [0] entry is the same as the 'unused' [1] entry).
Author: Mark Wutzke <mark.wutzke@alliedtelesis.co.nz>
git-svn-id: http://svn.fedorahosted.org/svn/corosync/trunk@1659 fd59a12c-fef9-0310-b244-a6a79926bd2f
Cause:
As part of its exit procedure, ais cancels its worker thread then manually
processes any outstanding items that were still in the worker thread's queue.
The worker thread has a low priority so normally it does not execute any
further before ais finishes exiting, but if the main thread's exiting is
delayed for any reason, there is a chance the worker thread could execute and
try to process items which have already been processed and freed by the main
thread - often leading to the worker thread seeing NULL data and ultimately
causing a segmentation fault.
Fix:
Modified worker_thread_group_exit() so it does a pthread_join() after the
pthread_cancel() call, so that the worker thread always shuts down cleanly
before the main thread does its cleanup.
Author: Author: Mark McKinstry <mark.mckinstry@alliedtelesis.co.nz>
git-svn-id: http://svn.fedorahosted.org/svn/corosync/trunk@1658 fd59a12c-fef9-0310-b244-a6a79926bd2f
Display strings safely, even if they are invalid, e.g. data in a received
message was corrupted.
Author: Tim Beale <tim.beale@alliedtelesis.co.nz>
git-svn-id: http://svn.fedorahosted.org/svn/corosync/trunk@1657 fd59a12c-fef9-0310-b244-a6a79926bd2f
This patch makes it possible to override the following #defines:
MESSAGE_SIZE_MAX
MESSAGE_QUEUE_MAX
SIZEQUEUE
FLOW_CONTROL_ENTRIES_ENABLE
If MESSAGE_SIZE_MAX is defined as 1024*64 (64K) and
MESSAGE_QUEUE_MAX defined as 512 you can change corosync's
memory footprint from ~48M to ~8M
So if you define MESSAGE_QUEUE_MAX, the queue size will
not be based on the message size any more.
To use this define the defines on the command line:
make CFLAGS="-DMESSAGE_SIZE_MAX=1024*64 -DMESSAGE_QUEUE_MAX=512"
git-svn-id: http://svn.fedorahosted.org/svn/corosync/trunk@1656 fd59a12c-fef9-0310-b244-a6a79926bd2f
Use a 2 phase "commit" operation:
1) Invoke verifyconfig that should catch the errors before the reload operation
2) Invoke reloadconfig that performs the operation and should _never_ fail
Implementation note: if step 2 fails, there is no fall back at the moment.
Fix the IPC table for confdb:
MESSAGE_REQ_CONFDB_XPATH_EVAL_EXPRESSION = 12 was added to include/ipc_confdb.h
without an associated call. Thanks Chrissie for spotting this.
git-svn-id: http://svn.fedorahosted.org/svn/corosync/trunk@1629 fd59a12c-fef9-0310-b244-a6a79926bd2f
logsys_config_priority_set lacked a check for logsys_single_id that
left logsys_subsys_id uninitialized.
git-svn-id: http://svn.fedorahosted.org/svn/corosync/trunk@1624 fd59a12c-fef9-0310-b244-a6a79926bd2f
This call causes a complete list of active groups and their
membership lists to be sent to a callback function.
git-svn-id: http://svn.fedorahosted.org/svn/corosync/trunk@1571 fd59a12c-fef9-0310-b244-a6a79926bd2f
LOG_MODE_SHORT_FILELINE strips the leading path from the file name in
a debug output. LOG_MODE_NOSUBSYS removes a global variable and provides
the ability to set or unset this via config_mode_set.
git-svn-id: http://svn.fedorahosted.org/svn/corosync/trunk@1570 fd59a12c-fef9-0310-b244-a6a79926bd2f
If the operation is succesful there is no need to set error_string. If error_string is not set, don't try to access it or we crash.
At the same time perform the same check in libconfdb when we receive the reply.
Fabio
git-svn-id: http://svn.fedorahosted.org/svn/corosync/trunk@1569 fd59a12c-fef9-0310-b244-a6a79926bd2f
