1  Mnesia Release Notes

1 Mnesia Release Notes

This document describes the changes made to the Mnesia system from version to version. The intention of this document is to list all incompatibilities as well as all enhancements and bugfixes for every release of Mnesia. Each release of Mnesia thus constitutes one section in this document. The title of each section is the version number of Mnesia.

  • Mnesia could crash during startup if del_table_copy/2 and add_table_copy/3 was invoked when the table was loading.

    Own Id: OTP-19076 Aux Id: ERIERL-1073

  • Document mnesia:foldl/4 and mnesia:foldr/4.

    Own Id: OTP-18798

  • mnesia:add_table_copy/3 no longer fails with reason system_limit when the node is starting.

    Own Id: OTP-18850

  • Restore recreate of disc_only tables could crash if they had an index.

    Own Id: OTP-18843 Aux Id: GH-7766

  • Do not delete old backup file if the new backup fails.

    Own Id: OTP-18711 Aux Id: ERIERL-963

  • Added debug statistics for active transactions.

    Own Id: OTP-18309 Aux Id: PR-6377

  • The implementation has been fixed to use proc_lib:init_fail/2,3 where appropriate, instead of proc_lib:init_ack/1,2.

    *** POTENTIAL INCOMPATIBILITY ***

    Own Id: OTP-18490 Aux Id: OTP-18471, GH-6339, PR-6843

  • mnesia:add_table_copy/3 no longer fails with reason system_limit when the node is starting.

    Own Id: OTP-18850

  • Do not delete old backup file if the new backup fails.

    Own Id: OTP-18711 Aux Id: ERIERL-963

  • Improved consistency for dirty writes when a table was added with add_table_copy/3.

    Fixed a problem with sticky write, which could lead to inconsistent data.

    Own Id: OTP-18412

  • Replace size/1 with either tuple_size/1 or byte_size/1

    The size/1 BIF is not optimized by the JIT, and its use can result in worse types for Dialyzer.

    When one knows that the value being tested must be a tuple, tuple_size/1 should always be preferred.

    When one knows that the value being tested must be a binary, byte_size/1 should be preferred. However, byte_size/1 also accepts a bitstring (rounding up size to a whole number of bytes), so one must make sure that the call to byte_size/ is preceded by a call to is_binary/1 to ensure that bitstrings are rejected. Note that the compiler removes redundant calls to is_binary/1, so if one is not sure whether previous code had made sure that the argument is a binary, it does not harm to add an is_binary/1 test immediately before the call to byte_size/1.

    Own Id: OTP-18432 Aux Id: GH-6672,PR-6793,PR-6784,PR-6787,PR-6785,PR-6682,PR-6800,PR-6797,PR-6798,PR-6799,PR-6796,PR-6813,PR-6671,PR-6673,PR-6684,PR-6694,GH-6677,PR-6696,PR-6670,PR-6674

  • Fixed crash which could happen during startup if too many decisions where sent from remote nodes.

    Own Id: OTP-18319 Aux Id: ERIERL-875

  • Don't fill the logs if mnesia can't connect to all nodes, due to partitioned network.

    Own Id: OTP-18288 Aux Id: ERIERL-868

  • Fixed add_table_copy which could leave a table lock if the receiving node went down during the operation.

    Own Id: OTP-18128 Aux Id: PR-6013

  • Documentation fixes.

    Own Id: OTP-17930

  • mnesia:add_table_copy/3 no longer fails with reason system_limit when the node is starting.

    Own Id: OTP-18850

  • Do not delete old backup file if the new backup fails.

    Own Id: OTP-18711 Aux Id: ERIERL-963

  • Don't fill the logs if mnesia can't connect to all nodes, due to partitioned network.

    Own Id: OTP-18288 Aux Id: ERIERL-868

  • Fixed crash which could happen during startup if too many decisions where sent from remote nodes.

    Own Id: OTP-18319 Aux Id: ERIERL-875

  • Fixed add_table_copy which could leave a table lock if the receiving node went down during the operation.

    Own Id: OTP-18128 Aux Id: PR-6013

  • Fixed mnesia:add_table_copy/3 so that calling it when mnesia started on another node does not fail or cause hanging nodes.

    Own Id: OTP-18056

  • Optimize locker to handle many read locks on the same record.

    Own Id: OTP-17973 Aux Id: ERIERL-772

  • Reduce the number of locks taken during table copying, should reduce the startup time on large systems.

    Own Id: OTP-17656 Aux Id: ERIERL-688

  • Documentation and minor code cleanup.

    Own Id: OTP-17727

  • Fixed that index keys was deleted for set tables when mnesia:delete_object/1 tried to delete a non-existing record.

    Own Id: OTP-17564 Aux Id: GH-5040

  • Optimized table loading and added max_transfer_size configuration parameter.

    Own Id: OTP-17508

  • Suppression of deprecation warnings has been added to the source files of the Mnesia application.

    Own Id: OTP-17217

  • Fixed that the backend plugin initialization is done only once.

    Own Id: OTP-17294 Aux Id: GH-4525 PR-4674

  • Fixed the type spec for disc_only_copies.

    Own Id: OTP-17249 Aux Id: PR-4578

  • Do not crash in mnesia:change_config/2 if mnesia is stopping or starting.

    Own Id: OTP-17274 Aux Id: GH-4616

  • Optimized table loading time for tables that are updated during the loading.

    Own Id: OTP-17271 Aux Id: PR-4575

  • Avoid potential performance issue, if the input queue to mnesia_tm is long.

    Own Id: OTP-17066 Aux Id: PR-2889

  • FIx mnesia delete object handling in transaction storage. In a transaction mnesia:read/1 could indicate that exiting objects did not exist after another object was deleted.

    Own Id: OTP-16782 Aux Id: PR-2663

  • Fixed crash during startup, which could happen if a table was deleted on another node.

    Own Id: OTP-16815 Aux Id: ERIERL-500

  • Make mnesia:create_table/2 return correct badarg value.

    Own Id: OTP-16072 Aux Id: PR-2320

  • Fixed a bug where mnesia was sometimes not waiting during start for a commit decision on asymmetric transactions.

    Own Id: OTP-16634 Aux Id: PR-2610 ERL-1227

  • Remove usage and documentation of old requests of the I/O-protocol.

    Own Id: OTP-15695

  • Avoid using rpc calls to do table reads, which will reduce the load on rpc server and improve performance.

    Own Id: OTP-16189

  • Fixed crash during startup, which could happen if a table was deleted on another node.

    Own Id: OTP-16815 Aux Id: ERIERL-500

  • Fixed a timing issue in uninstall fallback functionality.

    Own Id: OTP-16468 Aux Id: ERL-1151

  • Fixed mnesia crash which could happen when trying to recover from failures in transactions containing sticky_locks.

    Own Id: OTP-16286 Aux Id: ERL-1077

  • Fixed mnesia index issue. Could happen when updating records with a index plugin backend.

    Own Id: OTP-16291 Aux Id: ERL-1091

  • mnesia:add_table_copy/3 could cause a deadlock if called when a new node was starting.

    Own Id: OTP-15933 Aux Id: ERL-872

  • Transactions with sticky locks could with async_asym transactions be committed in the wrong order, since asym transaction are spawned on the remote nodes.

    To fix this bug the communication protocol between mnesia nodes had to be updated, thus mnesia will no longer be able to connect to nodes earlier than mnesia-4.14 , OTP-19.0.

    *** POTENTIAL INCOMPATIBILITY ***

    Own Id: OTP-15979 Aux Id: ERL-768

  • Optimize mnesia:read/1 if data have been written in the same transaction.

    Own Id: OTP-15550 Aux Id: PR-2029

  • Fixed bugs in table index plugin handling.

    Own Id: OTP-15689 Aux Id: PR-1695 ERL-556

  • Optimized dumping of tables with plugin backends.

    Own Id: OTP-15588 Aux Id: PR-2102

  • Include stacktrace in exception if a dirty activity errors, thus if user have matched on the error thrown it may not match any more.

    *** POTENTIAL INCOMPATIBILITY ***

    Own Id: OTP-15804 Aux Id: PR-2216

  • Avoid overload warnings caused by a race condition.

    Own Id: OTP-15619 Aux Id: ERIERL-310

  • Fixed type spec for mnesia:change_config/2.

    Own Id: OTP-15201 Aux Id: PR-1881

  • When master node is set do not force a load from ram_copies replica when there are no available disc_copies, since that would load an empty table. Wait until a disk replica is available or until user explicitly force_loads the table.

    Own Id: OTP-15221 Aux Id: ERIERL-217

  • Allow to add replicas even if all other replicas are down when the other replicas are not stored on disk.

    Own Id: OTP-15226 Aux Id: ERIERL-221

  • Fixed mnesia:delete_object/1 bug, where delete_object was deleting the record if it was written in the same transaction even if it was written to a different value.

    Own Id: OTP-15231 Aux Id: PR-1858

  • Fixed a bug where the bag table index data was not deleted when objects were deleted.

    Own Id: OTP-15243

  • Calls to erlang:get_stacktrace() are removed.

    Own Id: OTP-14861

  • When master node is set do not force a load from ram_copies replica when there are no available disc_copies, since that would load an empty table. Wait until a disk replica is available or until user explicitly force_loads the table.

    Own Id: OTP-15221 Aux Id: ERIERL-217

  • Allow to add replicas even if all other replicas are down when the other replicase are not stored on disk.

    Own Id: OTP-15226 Aux Id: ERIERL-221

  • Removed a quadratic behavior in startup. This change implies that backend plugins (if used) must be set when the schema is created or via configuration parameters before mnesia is started.

    Own Id: OTP-14829 Aux Id: ERIERL-84

  • Bad timing could crash mnesia after a checkpoint was deactivated and reactivated with the same checkpoint name on different tables.

    Own Id: OTP-14841 Aux Id: ERIERL-113

  • Fix backup error handling, the real failure reason was not returned.

    Own Id: OTP-14776 Aux Id: ERIERL-103

  • General Unicode improvements.

    Own Id: OTP-14462

  • Removed the wrapping of select continuations in extension plugin handling. This might require the user to rewrite user backend plugin if used.

    *** POTENTIAL INCOMPATIBILITY ***

    Own Id: OTP-14039

  • Fixed crash in checkpoint handling when table was deleted during backup.

    Own Id: OTP-14167

  • A continuation returned by mnesia:select/[14] should be reusable in different, non-transactional activities.

    Own Id: OTP-13944 Aux Id: PR-1184

  • Fixed crash when calling block_table multiple times. Could happen when having locks for a long time and restarting mnesia.

    Own Id: OTP-13970 Aux Id: Seq-13198

  • Change mnesia_tm process to have off-heap messages since mnesia_tm can be the receiver of many non-synchronized message from other nodes.

    Own Id: OTP-14074

  • Correct some minor documentation issues.

    Own Id: OTP-13891

  • Added experimental external backend plugin api. This adds the possibility for the user to write other storage backends for data, for example by using shared memory or ram-cached disk storage.

    The plugin api may change in future versions after being battle tested.

    Own Id: OTP-13058

  • Mnesia transactions could hang while waiting on a response from a node who had stopped.

    Own Id: OTP-13423

  • Avoid deadlock possibility in mnesia:del_table_copy/2

    Own Id: OTP-13284

  • Fixed a process and file descriptor leak in mnesia:restore/2.

    Own Id: OTP-13025 Aux Id: seq12957

  • Improved index updates to avoid a timing glitch in dirty_index_read.

    Own Id: OTP-12972

  • Mnesia's dirty functions did not always exit with {aborted, Reason} as documented when an error occurred.

    Own Id: OTP-12714

  • Consider file descriptors limits (emfile) as a fatal error and do not delete log files. Previously the error was seen as a corrupted disk and the log files deleted which caused data loss.

    Own Id: OTP-12807

  • Make Mnesia DCD dump behavior at start up optional, when turned off mnesia loads large disc_copies tables faster.

    Own Id: OTP-12481

  • Fixed race condition in protocol negotiation.

    Own Id: OTP-12473

  • Grammar corrections. (Thanks to Derek Brown)

    Own Id: OTP-12400

  • Fixed a spelling mistake in mnesia documentation.

    Own Id: OTP-12278

  • Matching data with mnesia:match_object/1 did not work as expected in some cases, when data was written in the same transaction before the matching was invoked.

    Own Id: OTP-12304 Aux Id: Seq12745

  • Various logging fixes, including: Add run queue index to the process dump in crash dumps.
    Add thread index to enomem slogan when crashing.
    Remove error logger message for sending messages to old instances of the same node.

    Own Id: OTP-12115

  • Fixed a race which could make create_table fail if a node was going down during the transaction.

    Own Id: OTP-12124 Aux Id: seq12694

  • Force load table could hang when a node went away during start up.

    Own Id: OTP-11948 Aux Id: seq12585

  • The time for inserting locks for a transaction with large number of locks is reduced significantly.

    Own Id: OTP-11981

  • Some local implementations of removing the last element from a list are replaced by lists:droplast/1. Note that this requires at least stdlib-2.0, which is the stdlib version delivered in OTP 17.0. (Thanks to Hans Svensson)

    Own Id: OTP-11678

  • Application upgrade (appup) files are corrected for the following applications:

    asn1, common_test, compiler, crypto, debugger, dialyzer, edoc, eldap, erl_docgen, et, eunit, gs, hipe, inets, observer, odbc, os_mon, otp_mibs, parsetools, percept, public_key, reltool, runtime_tools, ssh, syntax_tools, test_server, tools, typer, webtool, wx, xmerl

    A new test utility for testing appup files is added to test_server. This is now used by most applications in OTP.

    (Thanks to Tobias Schlager)

    Own Id: OTP-11744

  • To prevent a race condition if there is a short communication problem when node-down and node-up events are received. They are now stored and later checked if the node came up just before mnesia flagged the node as down. (Thanks to Jonas Falkevik )

    Own Id: OTP-11497

  • Added mnesia:sync_log/0 to explicit sync mnesias transaction log.

    Own Id: OTP-11729

  • Fixed a race in mnesia which could cause hanging transaction when sticky locks had been used. Thanks janchochol.

    Own Id: OTP-11375

  • Fixed dirty_update_counter which could return ok, thanks Anton Ryabkov.

    Own Id: OTP-11485

  • Fix timing issues in checkpoint creation.

    Own Id: OTP-10957

  • Fixed a problem where the fallback BUP file is removed when calling mnesia:uninstall_fallback and mnesia is not started.

    Own Id: OTP-11241

  • If mnesia:clear_table/2 was called during a table load on that table, the schema record was written to the table instead of clearing table.

    Own Id: OTP-11030 Aux Id: seq12267

  • Optimize index creation for Mnesia set tables. Thanks to Nick Marino.

    Own Id: OTP-11103

  • Use chained send_after instead of send_interval, to make decrease the number of messages sent after a sleep (Thanks to James Wheare)

    Own Id: OTP-10636

  • Fix format of mnesia overload message (Thanks to Ahmed Omar)

    Own Id: OTP-10639

  • Added a general framework for executing benchmarks of Erlang/OTP. Benchmarks for the Erlang VM and mnesia have been incorporated in the framework.

    For details about how to add more benchmarks see $ERL_TOP/HOWTO/BENCHMARKS.md in the source distribution.

    Own Id: OTP-10156

  • Where necessary a comment stating encoding has been added to Erlang files. The comment is meant to be removed in Erlang/OTP R17B when UTF-8 becomes the default encoding.

    Own Id: OTP-10630

  • Remove support for the query keyword and query expressions. Thanks to Loïc Hoguin.

    Own Id: OTP-10729

  • Add tests showing that trying to delete non-existing object may corrupt the

    In case of bag tables, trying to delete a non-existing object leads to the index becoming corrupt. This happens if the non-existing object we try to delete happens to share its key and index field value with a single existing object in the table. Result: The index entry corresponding to the existing object is removed.

    Prevent index from being corrupted if a nonexistent item is deleted

    We have to ensure that we actually delete the last object with a given (key, index) pair before removing the index. Thanks to Bartlomiej Puzon

    Own Id: OTP-10220

  • Returns the same value for mnesia_loader:disc_load_table/2 as mnesia_loader:net_load_table/4 if a table copy cannot be found. (Thanks to Uwe Dauernheim)

    Own Id: OTP-10015

  • Improved table lock algorithm.

    Own Id: OTP-9890

  • Reduce calls to phash in key_to_frag_number

    Original code calls phash 1..2 times, based on which fragment the hashed key targets and how many fragments exist. New code always calls phash only once.

    Add mnesia_frag_hash test (Thanks to Philip Robinson)

    Own Id: OTP-9722

  • Fixed a sticky lock bug which caused mnesia:read(Tab, Key, write) return undefined.

    Own Id: OTP-9786

  • Use the synchronous log_terms instead of alog_terms in mnesia_log:ets2dcd()

    This avoids the situation where mnesia could dump a very large ets table in its entirety into the message queue of the disk_log process, causing memory blowup and choking the disk logger. (Thanks to Richard Carlsson)

    Own Id: OTP-9804

  • Implemented a new option to mnesia:create_table/2 which allows the user to assign 'ets' and 'dets' options not available in mnesia.

    Own Id: OTP-8970

  • Fix deadlock in mnesia:del_table_copy/2.

    Own Id: OTP-9689 Aux Id: seq11927

  • Allow schema operations when using different mnesia versions.

    Own Id: OTP-9657 Aux Id: seq11926

  • Fix protocol issues. Mnesia-4.4.19 could not communicate with to older nodes.

    Own Id: OTP-9473

  • XML files have been corrected.

    Own Id: OTP-9550 Aux Id: OTP-9541

  • Dump the log even if no transactions have been invoked on local node, otherwise the log will grow forever with decisions from the other nodes who have tables on disk. Thanks Marek Majkowski.

    Own Id: OTP-9551

  • Use dedicated api for clear_table, i.e. instead of match_delete use delete_all_objects. Thanks KukHyun Lee.

    Own Id: OTP-9558

  • Mnesia could crash if mnesia:add_table_index/2 was invoked before the table was loaded on all nodes.

    Own Id: OTP-9285 Aux Id: seq11844

  • Add {majority, boolean()} per-table option.

    With {majority, true} set for a table, write transactions will abort if they cannot commit to a majority of the nodes that have a copy of the table. Currently, the implementation hooks into the prepare_commit, and forces an asymmetric transaction if the commit set affects any table with the majority flag set. In the commit itself, the transaction will abort if it cannot satisfy the majority requirement for all tables involved in the transaction.(Thanks to Ulf Wiger)

    Own Id: OTP-9304

  • Call chmod without the "-f" flag

    "-f" is a non-standard chmod option which at least SGI IRIX and HP UX do not support. As the only effect of the "-f" flag is to suppress warning messages, it can be safely omitted. (Thanks to Holger Weiß)

    Own Id: OTP-9170

  • Mnesia sometimes failed to update meta-information in large systems, which could cause table content to be inconsistent between nodes.

    Own Id: OTP-9186 Aux Id: seq11728

  • Calling mnesia:first/1 on empty fragmented table works. Thanks Magnus Henoch.

    Own Id: OTP-9108

  • If Mnesia detects that the network is not fully connected during start, Mnesia will not start until all nodes are reachable.

    Own Id: OTP-9115 Aux Id: seq-11728

  • Fix issues reported by dialyzer.

    Own Id: OTP-9107

  • Sometimes a 'log_header' record was added to tables when invoking mnesia:restore/2 with the option 'recreate_tables'. Thanks Vance Shipley.

    Own Id: OTP-8960

  • Compiler warnings were eliminated.

    Own Id: OTP-8855

  • Eliminated warnings for auto-imported BIF clashes.

    Own Id: OTP-8840

  • Added mnesia:subscribe(activity) contributed by Bernard Duggan.

    Own Id: OTP-8519

  • Transactions could be left hanging if a node went down when invoking mnesia:sync_transaction/[1,2]. Thanks Igor Ribeiro Sucupira.

    Own Id: OTP-8402

  • Igor Ribeiro Sucupira added the option to compress data when copying tables between Mnesia nodes.

    Own Id: OTP-8406

  • The documentation is now built with open source tools (xsltproc and fop) that exists on most platforms. One visible change is that the frames are removed.

    Own Id: OTP-8250

  • Fixed duplicate results with mnesia:index_read() on ordered_set tables. Reported by Sam Bobroff.

    Fixed locking in mnesia:index_read() which now grabs a read table lock to ensure correctness, this may slow down the operation or block other processes trying to reach the same table.

    Calling mnesia:dump_log() could crash mnesia, Reported by Igor Ribeiro Sucupira.

    Own Id: OTP-8074

  • Mnesia crashed if a qlc query was running inside a transaction when mnesia stopped at another node. Thanks Teemu Antti-Poika.

    Own Id: OTP-7968

  • Mnesia could crash when loading local_content tables.

    Own Id: OTP-8002 Aux Id: seq11277

  • Minor (smp) optimizations.

    Own Id: OTP-7928

  • mnesia:clear_table/1 crashed instead of returning {aborted,..} if it was called inside a transaction.

    Own Id: OTP-7911

  • With bad timing several api functions could return or exit with a bad error message when mnesia was shutting down.

    Own Id: OTP-7753 Aux Id: seq11179

  • mnesia:clear_table/1 cleared all nodes table content even if the table was local_content only type.

    Own Id: OTP-7835

  • Disallowed match patterns ('_', and '$n') as argument to mnesia:delete_object/1 and friends.

    Own Id: OTP-7524

  • Introduced a few new functions in Mnesia: mnesia:read/2, mnesia:first/3, mnesia:last/3, mnesia:prev/4, mnesia:next/4, mnesia_frag:first/1, mnesia_frag:last/1, mnesia_frag:prev/2, mnesia_frag:next/2.

    Own Id: OTP-7625

  • mnesia:restore/2 aborted if a EXIT message appeared in the client message queue.

    Own Id: OTP-7585 Aux Id: seq11046

  • mnesia:clear_table/1 does not require that all replicas of the table are available anymore.

    Own Id: OTP-7466 Aux Id: seq11015

  • Mnesia did not garbage collect transaction decisions on disk based nodes if no transactions where made on the local node.

    Own Id: OTP-7419

  • Table referred to by foreign key did not have node_pool properly cleaned up when a node was removed from the schema. Thanks Paul Mineiro.

    Own Id: OTP-7340

  • Mnesia crashed and generated a core dump if a schema_transaction was running when mnesia stopped.

    Own Id: OTP-7378 Aux Id: seq10964

  • It is now possible to delete a db node even when other disk resident nodes are down. Thanks Paul Mineiro.

    Own Id: OTP-7383

  • Sticky locks could lead to hanging transactions.

    Own Id: OTP-7205 Aux Id: seq10793

  • mnesia:snmp_get_next_index/2 didn't work with partial index keys. Argument checking is now done according to documentation, in functions mnesia:snmp_get_row/2, mnesia:snmp_get_mnesia_key/2 and mnesia:snmp_get_next_index/2. These functions now require that RowIndex is a list.

    *** POTENTIAL INCOMPATIBILITY ***

    Own Id: OTP-7208

  • Snmp index tables was not initialized correctly in mnesia-4.4.

    Own Id: OTP-7170 Aux Id: seq10870

  • Rearranging fragmented tables is an O(N^2) operation.

    Own Id: OTP-6300

  • Mnesia ignored the module argument to mnesia:restore/2. Thanks Paul Minerio.

    Own Id: OTP-6981

  • Mnesia's snmp operations snmp_get_row/2, snmp_get_next_index/2 and snmp_get_mnesia_key/2 have been made context aware, i.e. inside a transaction they will compensate for table updates made in earlier in the same transaction. This might cause a performance drop if a lot of updates have been made before the invocation of these functions.

    *** POTENTIAL INCOMPATIBILITY ***

    Own Id: OTP-6856 Aux Id: seq10671

  • Introduced erlang:phash/2 as new default for fragmented tables. Already existing tables will continue to use whatever hash function they where using.

    Own Id: OTP-6923

  • Introduced mnesia:is_transaction/0.

    Own Id: OTP-6995 Aux Id: seq10812

  • Rearranging fragmented tables is an O(N^2) operation.

    Own Id: OTP-6300