The annotated tag, ctdb-2.5.4 has been created at fef22f279b812adacbf8cfff134d969219a035f1 (tag) tagging 70c7ef023730d8344ca4afde2c94634dd541101f (commit) replaces ctdb-2.5.3 tagged by Amitay Isaacs on Fri Sep 26 14:53:56 2014 +1000
- Log ----------------------------------------------------------------- new version 2.5.4 Amitay Isaacs (97): daemon: Always update database priority cluster wide daemon: Add control CTDB_CONTROL_DB_DETACH client: Add client code to detach a database tools/ctdb: Add ctdb detach command to detach databases doc: Add "ctdb detach" and update "ctdb attach" doc: Remove commands that have been deleted tests: Add a simple test for "ctdb detach" daemon: Do not allow database detach if AllowClientDBAttach=1 tools/ctdb: Detach databases only if all nodes disallow client access daemon: Talloc tdb_wrap off ctdb_db_context client: Talloc tdb_wrap off ctdb_db_context tools/ctdb: Unlock records before closing tdb database tests: Add test for re-attaching detached database recoverd: Detach database from recovery daemon pmda: Do not hardcode include paths ib: Make infiniband transport compile again common: No need to save previous scheduler priority common: Do not abort if restoring scheduling policy fails common: Changing scheduler policy does not require ctdb context common: Drop ctdb prefix from utility functions independent of ctdb common: Separate system utilties that are ctdb independent common: Separate more system utilities that are independent of ctdb daemon: No need to block SIGPIPE at startup common: Remove unused functions daemon: Rename block_signal to ignore_signal common: Keep debug level related functions with logging code daemon: Instead of passing ctdb context, pass valgrinding boolean daemon: Rename ctdb_lockdown_memory to lockdown_memory common: Move lockdown_memory to system utilities daemon: Rename ctdb_mkdir_p_or_die to mkdir_p_or_die common: Move mkdir_p_or_die to system utilities logging: Split ringbuffer handling code from ctdb_collect_log logging: Move controls handling functions from common to server daemon: Remove duplicate code from helper binaries build: Remove duplicate library dependency daemon: Do not complain if node is inactive and db is not attached daemon: Reset scheduler policy for helper processes recoverd: No need to set ctdbd_pid again Revert "It was possible for ->recovery_mode to get out of sync with the new three db priorities in such a way that" tools: There is no need for forcing a recovery recoverd: Set recovery mode before freezing databases daemon: Do not thaw databases if recovery is active tests: Check that ctdb wipedb cleans the database tests: Add a test for ctdb restoredb daemon: Use false instead of 0 for boolean arguments daemon: Simplify code a bit daemon: Enable robust mutexes only if TDB_MUTEX_LOCKING is defined daemon: Support per-node robust mutex feature ltdb: Use tdb_null instead of zeroing TDB_DATA variable vacuum: Fix talloc hierarchy in delete_marshall_traverse vacuum: If talloc_realloc fails, terminate traverse util: Fix nonempty line endings util: Refactor ctdb_marshall_record util: Refactor record marshalling routines to avoid extra talloc vacuum: Use ctdb_marshall_add to add a record to marshall buffer vacuum: Use existing function ctdb_marshall_finish locking: Fix the lock_type_str corresponding to LOCK_ALLDB locking: Remove unused structure members locking: Remove multiple lock requests per lock context (part 1) locking: Remove multiple lock requests per lock context (part 2) locking: Schedule the next possible lock based on per-db limit locking: Remove unused function find_lock_context() locking: Allocate lock request soon after allocating lock context locking: Add new tunable LockProcessesPerDB locking: Increase number of lock processes per database to 200 locking: Remove unused variable lock_num_pending locking: Add DB lock requests to head of the pending queue locking: Update ctdb statistics for all lock types locking: Decrement pending statistics when lock is scheduled locking: Simplify check for locks on record or database locking: Update a comment locking: Add per database queues for pending and active lock requests readonly: Add an early return to simplify code readonly: Do not abort if revoke of readonly record fails on a node locking: Run debug locks script only if the node is active locking: Talloc lock request from client specified context locking: Remove unused function ctdb_free_lock_request_context include: Remove declaration of non-existent function common: Refactor code to convert TDB_DATA key to aligned uint32 array daemon: Remove duplicate code with refactored function daemon: Defer all calls when processing dmaster packets locking: Do not reset real-time priority for lock helpers call: Drop all deferred requests from older generation daemon: Increment pending calls statistics correctly locking: Set real-time priority for lock helpers daemon: Make sure ctdb runs with real-time priority tests: Do not expect real-time priority when running local daemons daemon: Decrement pending calls statistics when calls are deferred doc: Add ctdb-statistics manual page doc: Add reference to new magepage ctdb-statistics scripts: Fix a typo scripts: Do not export variables if they are not set doc: Update NEWS locking: Reset ttimer before doing an early return scripts: Fix the regular expresssion for parsing /proc/locks recoverd: If obtaining recovery lock fails, try again daemon: Fix the usage for lock helper Martin Schwenke (40): tools-ctdb: Drop disconnected nodes when filtering by capability eventscripts: CTDB_NATGW_PUBLIC_* optional on slave-only nodes tests: Update "ctdb detach" test eventscripts: Don't check if $iface is empty eventscripts: Update comment in 11.routing eventscripts: Add check for invalid policy routing configuration daemon: Move a ZERO_STRUCT() to a better place tools-ctdb: scriptstatus should not count disabled scripts tools-ctdb: Make natgwlist and lvsmaster more resilient eventscripts: New configuration variable CTDB_GANESHA_REC_SUBDIR common: Do not dynamically allocate memory when not necessary common: No dynamic memory allocation in mkdir_p() tests: Try harder to avoid failures due to repeated recoveries daemon: Debugging for tickle updates eventscripts: Ensure $GANRECDIR points to configured subdirectory scripts: Always print footer when debugging hung script tests: Fix racy test for debugging hung scripts daemon: Exit if setting the session ID fails daemon: Check PID in ctdb_remove_pidfile(), not unreliable flag daemon: Remove ctdbd_pid global variable scripts: Support NFS on RHEL7 with systemd recoverd: Log a message when releasing the recovery lock recoverd: Don't say "Election timed out" common: Use SCHED_RESET_ON_FORK when setting SCHED_FIFO recoverd: Gently abort recovery when election is underway eventscripts: Remove NAT gateway "monitor" event tests: Add another LCP2 takeover test eventscripts: Remove unused argument to natgw_ensure_master() eventscripts: Remove special case for virtio_net doc: Fix default database directories in ctdbd.1 replace: Remove unused item returned by FAILED() tools: Factor out new function find_node_xpnn() from control_xpnn() tools: Be more helpful when CTDB CLI tool is run on unconfigured node common: Copy functions sys_read() and sys_write() from source3 Use sys_read() and sys_write() to ensure correct signal interaction util: Fix warning about ignored result from system() daemon: Fix some strict-aliasing warnings locking: Add argc parameter to lock_helper_args() locking: Fork lock helper with vfork_with_logging() util: Log an error if there is no way to set scheduler Michael Adam (14): server: fix DEBUG message for wrong event script options. vacuum: adapt debug message for repacking. vacuum: remove a comment in ctdb_vacuum_and_repack_db() vacuum: use plain tdb_repack() instead of ctdb_repack_tdb() vacuum: remove now unused ctdb_repack_tdb(). vacuum: remove a superfluous and misleading comment vacuum: untangle assignmend and check for return of tdb_repack() vacuum: remove a superfluous comment. vacuum: remove vacuum limit from vdata - not used vacuum: move init of vdata into init_vdata funcion vacuum: remove now unused talloc ctx argument from ctdb_vacuum_db() vacuum: add missing return to ctdb_vacuum_traverse_db() error path. vacuum: always run freelist_size again vacuum: stop vacuuming when the first delete_list traverse fails. Volker Lendecke (5): daemon: Allow flag TDB_MUTEX_LOCKING to pass into db_attach common: Avoid a talloc in ctdb_queue_send common: Fix verbose_memory_names locking: TALLOC_FREE copes with NULL locking: Simplify ctdb_find_lock_context() ----------------------------------------------------------------------- -- CTDB repository