7 * 2 new configuration variables for 60.nfs eventscript:
9 - CTDB_MONITOR_NFS_THREAD_COUNT
10 - CTDB_NFS_DUMP_STUCK_THREADS
12 See ctdb.sysconfig for details.
17 * ctdb tool should not exit from a retry loop if a control times out
18 (e.g. under high load). This simple fix will stop an exit from the
19 retry loop on any error.
21 Important internal changes
22 --------------------------
33 * The "stopped" event has been removed.
35 The "ipreallocated" event is now run when a node is stopped. Use
36 this instead of "stopped".
38 * New --pidfile option for ctdbd, used by initscript
40 * The 60.nfs eventscript now uses configuration files in
41 /etc/ctdb/nfs-rpc-checks.d/ for timeouts and actions instead of
42 hardcoding them into the script.
44 * Notification handler scripts can now be dropped into /etc/ctdb/notify.d/.
46 * The NoIPTakeoverOnDisabled tunable has been renamed to
47 NoIPHostOnAllDisabled and now works properly when set on individual
50 * New ctdb subcommand "runstate" prints the current internal runstate.
51 Runstates are used for serialising startup.
56 * The Unix domain socket is now set to non-blocking after the
57 connection succeeds. This avoids connections failing with EAGAIN
58 and not being retried.
60 * Fetching from the log ringbuffer now succeeds if the buffer is full.
62 * Fix a severe recovery bug that can lead to data corruption for SMB clients.
64 * The statd-callout script now runs as root via sudo.
66 * "ctdb delip" no longer fails if it is unable to move the IP.
68 * A race in the ctdb tool's ipreallocate code was fixed. This fixes
69 potential bugs in the "disable", "enable", "stop", "continue",
70 "ban", "unban", "ipreallocate" and "sync" commands.
72 * The monitor cancellation code could sometimes hang indefinitely.
73 This could cause "ctdb stop" and "ctdb shutdown" to fail.
75 Important internal changes
76 --------------------------
78 * The socket I/O handling has been optimised to improve performance.
80 * IPs will not be assigned to nodes during CTDB initialisation. They
81 will only be assigned to nodes that are in the "running" runstate.
83 * Improved database locking code. One improvement is to use a
84 standalone locking helper executable - the avoids creating many
85 forked copies of ctdbd and potentially running a node out of memory.
87 * New control CTDB_CONTROL_IPREALLOCATED is now used to generate
88 "ipreallocated" events.
90 * Message handlers are now indexed, providing a significant
91 performance improvement.