April 2017 - discuss - lists.mariadb.org

[Maria-discuss] why stdout? -> [Note] /usr/libexec/mysqld (mysqld 10.0.20-MariaDB) starting as process 115096
by Reindl Harald 05 Jul '20

05 Jul '20

Aug 6 12:24:24 testserver mysqld: 150806 12:24:24 [Note] /usr/libexec/mysqld (mysqld 10.0.20-MariaDB) starting as process 115096 ... why is this now stdout instead mysql logs? on systemd based distributions that spits now to the global syslog and finally on machines restarting 2 instances every hour for backup reasons it annoys until 10.0.19 or 10.0.20 that was not this way

5 13

[Maria-discuss] Cross compile error
by Lucian Cristian 16 May '17

16 May '17

Hi I'm trying to cross compile (any 10.* versions) using musl and uClibc++ for a LEDE/Openwrt system and I have problems with these errors: staging_dir/target-x86_64_musl/usr/include/uClibc++/associative_base:444:8: error: 'void std::__base_associative<Key, ValueType, Compare, Allocator>::swap(std::__base_associative<Key, ValueType, Compare, Allocator>&) [with Key = dict_foreign_t*; ValueType = dict_foreign_t*; Compare = dict_foreign_compare; Allocator = ut_allocator<dict_foreign_t*>]' is protected void __base_associative<Key, ValueType, Compare, Allocator>::swap(__base_associative<Key, ValueType, Compare, Allocator>& m) ^ build_dir/target-x86_64_musl/mariadb-10.2.4/storage/innobase/dict/dict0dict.cc:1996:32: error: within this context table->foreign_set.swap(fk_set); ^ storage/innobase/CMakeFiles/innobase.dir/build.make:542: recipe for target 'storage/innobase/CMakeFiles/innobase.dir/dict/dict0dict.cc.o' failed staging_dir/target-x86_64_musl/usr/include/uClibc++/map: In instantiation of 'std::map<Key, T, Compare, Allocator>::reference std::map<Key, T, Compare, Allocator>::operator[](const key_type&) [with Key = const char*; T = dict_index_t*; Compare = ut_strcmp_functor; Allocator = ut_allocator<std::pair<const char*, dict_index_t*> >; std::map<Key, T, Compare, Allocator>::reference = std::pair<const char*, dict_index_t*>&; std::map<Key, T, Compare, Allocator>::key_type = const char*]': build_dir/target-x86_64_musl/mariadb-10.2.4/storage/innobase/dict/dict0stats.cc:2518:22: required from here staging_dir/target-x86_64_musl/usr/include/uClibc++/map:143:13: error: invalid initialization of reference of type 'std::map<const char*, dict_index_t*, ut_strcmp_functor, ut_allocator<std::pair<const char*, dict_index_t*> > >::reference {aka std::pair<const char*, dict_index_t*>&}' from expression of type 'dict_index_t*' return i->second; ^ storage/innobase/CMakeFiles/innobase.dir/build.make:614: recipe for target 'storage/innobase/CMakeFiles/innobase.dir/dict/dict0stats.cc.o' failed build_dir/target-x86_64_musl/mariadb-10.2.4/plugin/handler_socket/handlersocket/hstcpsvr.cpp: In constructor 'dena::hstcpsvr::hstcpsvr(const dena::config&)': build_dir/target-x86_64_musl/mariadb-10.2.4/plugin/handler_socket/handlersocket/hstcpsvr.cpp:92:37: error: invalid initialization of non-const reference of type 'std::auto_ptr<volatile dena::database_i>&' from an rvalue of type 'dena::database_ptr {aka std::auto_ptr<volatile dena::database_i>}' cshared.dbptr = database_i::create(c); ^ In file included from staging_dir/target-x86_64_musl/usr/include/uClibc++/vector:22:0, from build_dir/target-x86_64_musl/mariadb-10.2.4/plugin/handler_socket/handlersocket/hstcpsvr.cpp:12: staging_dir/target-x86_64_musl/usr/include/uClibc++/memory:137:12: note: initializing argument 1 of 'std::auto_ptr<T>& std::auto_ptr<T>::operator=(std::auto_ptr<T>&) [with T = volatile dena::database_i]' auto_ptr& operator=(auto_ptr& p) throw(){ ^ plugin/handler_socket/CMakeFiles/handlersocket.dir/build.make:134: recipe for target 'plugin/handler_socket/CMakeFiles/handlersocket.dir/handlersocket/hstcpsvr.cpp.o' failed make[5]: *** [plugin/handler_socket/CMakeFiles/handlersocket.dir/handlersocket/hstcpsvr.cpp.o] Error 1 this is the command line (cd /build_dir/target-x86_64_musl/mariadb-10.2.4; CFLAGS="-Os -pipe -fno-caller-saves -fno-plt -fhonour-copts -Wno-error=unused-but-set-variable -Wno-error=unused-result -iremap/build_dir/target-x86_64_musl/mariadb-10.2.4:mariadb-10.2.4 -Wformat -Werror=format-security -fstack-protector -D_FORTIFY_SOURCE=1 -Wl,-z,now -Wl,-z,relro -fpic " CXXFLAGS="-Os -pipe -fno-caller-saves -fno-plt -fhonour-copts -Wno-error=unused-but-set-variable -Wno-error=unused-result -iremap/build_dir/target-x86_64_musl/mariadb-10.2.4:mariadb-10.2.4 -Wformat -Werror=format-security -fstack-protector -D_FORTIFY_SOURCE=1 -Wl,-z,now -Wl,-z,relro -fpic " LDFLAGS="-L/staging_dir/target-x86_64_musl/usr/lib -L/staging_dir/target-x86_64_musl/lib -L/staging_dir/toolchain-x86_64_gcc-5.4.0_musl/usr/lib -L/staging_dir/toolchain-x86_64_gcc-5.4.0_musl/lib -znow -zrelro " cmake -DCMAKE_SYSTEM_NAME=Linux -DCMAKE_SYSTEM_VERSION=1 -DCMAKE_SYSTEM_PROCESSOR=x86_64 -DCMAKE_BUILD_TYPE=Release -DCMAKE_C_FLAGS_RELEASE="-DNDEBUG" -DCMAKE_CXX_FLAGS_RELEASE="-DNDEBUG" -DCMAKE_C_COMPILER="/staging_dir/toolchain-x86_64_gcc-5.4.0_musl/bin/x86_64-openwrt-linux-musl-gcc" -DCMAKE_C_COMPILER_ARG1="" -DCMAKE_CXX_COMPILER="/staging_dir/toolchain-x86_64_gcc-5.4.0_musl/bin/g++-uc" -DCMAKE_CXX_COMPILER_ARG1="" -DCMAKE_ASM_COMPILER="/staging_dir/toolchain-x86_64_gcc-5.4.0_musl/bin/x86_64-openwrt-linux-musl-gcc" -DCMAKE_ASM_COMPILER_ARG1="" -DCMAKE_EXE_LINKER_FLAGS:STRING="-L/staging_dir/target-x86_64_musl/usr/lib -L/staging_dir/target-x86_64_musl/lib -L/staging_dir/toolchain-x86_64_gcc-5.4.0_musl/usr/lib -L/staging_dir/toolchain-x86_64_gcc-5.4.0_musl/lib -znow -zrelro" -DCMAKE_MODULE_LINKER_FLAGS:STRING="-L/staging_dir/target-x86_64_musl/usr/lib -L/staging_dir/target-x86_64_musl/lib -L/staging_dir/toolchain-x86_64_gcc-5.4.0_musl/usr/lib -L/staging_dir/toolchain-x86_64_gcc-5.4.0_musl/lib -znow -zrelro -Wl,-Bsymbolic-functions" -DCMAKE_SHARED_LINKER_FLAGS:STRING="-L/staging_dir/target-x86_64_musl/usr/lib -L/staging_dir/target-x86_64_musl/lib -L/staging_dir/toolchain-x86_64_gcc-5.4.0_musl/usr/lib -L/staging_dir/toolchain-x86_64_gcc-5.4.0_musl/lib -znow -zrelro -Wl,-Bsymbolic-functions" -DCMAKE_AR="/staging_dir/toolchain-x86_64_gcc-5.4.0_musl/bin/x86_64-openwrt-linux-musl-gcc-ar" -DCMAKE_NM="/staging_dir/toolchain-x86_64_gcc-5.4.0_musl/bin/x86_64-openwrt-linux-musl-gcc-nm" -DCMAKE_RANLIB="/staging_dir/toolchain-x86_64_gcc-5.4.0_musl/bin/x86_64-openwrt-linux-musl-gcc-ranlib" -DCMAKE_FIND_ROOT_PATH="/staging_dir/target-x86_64_musl/usr;/staging_dir/toolchain-x86_64_gcc-5.4.0_musl" -DCMAKE_FIND_ROOT_PATH_MODE_PROGRAM=BOTH -DCMAKE_FIND_ROOT_PATH_MODE_LIBRARY=ONLY -DCMAKE_FIND_ROOT_PATH_MODE_INCLUDE=ONLY -DCMAKE_STRIP=: -DCMAKE_INSTALL_PREFIX=/usr -DDL_LIBRARY=/staging_dir/target-x86_64_musl -DCMAKE_PREFIX_PATH=/staging_dir/target-x86_64_musl -DCMAKE_SKIP_RPATH=TRUE -DCMAKE_SYSTEM_NAME=Linux -DSTACK_DIRECTION=-1 -DWITH_WSREP=ON -DWITH_INNODB_DISALLOW_WRITES=1 -DCONNECT_WITH_JDBC=OFF -DWITHOUT_TOKUDB=1 -DIMPORT_EXECUTABLES=/build_dir/hostpkg/mariadb-10.2.4/host/import_executables.cmake /build_dir/target-x86_64_musl/mariadb-10.2.4 ) can someone give me an advice ? Regards

2 2

Re: [Maria-discuss] MySQL Workbench Fork?
by Federico Razzoli 07 May '17

07 May '17

Well, clicking around, selecting things, switching to keyboards, and so on, can hardly be faster than writing an SQL statement - provided that you learn ALTER TABLE syntax. But, apart from that, I think that we mostly agree on everything except one thing: the meaning of enterprise. In an enterprise environment a DBA needs to automate everything and write scripts. Also, as I wrote before and the author of that post also says, GUI's are not reliable, you never know exactly what they do. That said, I suppose that GUIs can be useful for someone, especially for developers and hobbyists. Cheers, Federico -------------------------------------------- Gio 27/4/17, <info(a)smallbusinessconsultingexperts.com> ha scritto: Oggetto: Re: [Maria-discuss] MySQL Workbench Fork? A: maria-discuss(a)lists.launchpad.net, "Federico Razzoli" <federico_raz(a)yahoo.it> Data: Giovedì 27 Aprile 2017, 20:34 Sure, but don't confuse a post with a fact either. That's great he has an opinion that you share, I am sure I can find many more that say the opposite. What I got from his post is there are some things a gui cannot do, which is of course the case. So? The opposite is true also, if you add the word quickly. That does not make a point as universal true or factual. Should we abandon SNMP and other standards type network consoles (gui) since obviously we can do the same thing without them? You don't see the value is the gui network view? So, workbench feature of showing you a refreshing graphical status of your server isn't useful? It's not complete by any means, but it is useful, just like the connections screen showing running queries. Those are just examples which some will desire to pick on. But I think it's a reasonable view. My point is not guis are (always) better, they each have a place. On Apr 27, 2017, 1:23 PM -0500, Federico Razzoli <federico_raz(a)yahoo.it>, wrote: I've just seen this Quora answer, which lists some good reasons to avoid GUIs. There are others. Again: my point is not "GUI's shouldn't exist", it's just "please don't call them an enterprise software, because they are not". https://www.quora.com/Do-database-administrators-actually-think-that-GUIs-a…

8 8

[Maria-discuss] MySQL Workbench Fork?
by Zach Lym 05 May '17

05 May '17

I convinced my boss to use MariaDB for a new project but it appears that the workbench is incompatible with new features of MariaDB. I've looked at the alternatives listed on the website, but they are all either proprietary or Windows only. I just can't sell my boss on switching to MariaDB without workbench. Free software isn't free, but I think Workbench is a big driver of commercial and enterprise use. It's easy to kvetch from the outside, but I feel as though MariaDB has just lost some ground. Has management considered forking Workbench? Thank you, -Zach Lym

9 12

[Maria-discuss] lz4 innodb compression
by Reinis Rozitis 28 Apr '17

28 Apr '17

Hello, is lz4 support not compiled in the Mariadb rpm (Opensuse) packages or are there some runtime requirements or you have to rebuild from source? MariaDB> set global innodb_compression_algorithm=lz4; ERROR 1231 (42000): Variable 'innodb_compression_algorithm' can't be set to the value of 'lz4' while MariaDB> set global innodb_compression_algorithm=lzo; Query OK, 0 rows affected (0.00 sec) Server version: 10.2.5-MariaDB-log MariaDB Server / OpenSUSE Leap 42.2 / on system: lz4-1.4.0+svn124-3.3.x86_64 liblz4-1_4-1.4.0+svn124-3.3.x86_64 rr

3 4

[Maria-discuss] howto debug “Slave SQL: Error executing row event” and “Node consistency compromized, aborting…” and “Failed to apply trx: source”
by Lukas Lehner 28 Apr '17

28 Apr '17

Hi https://dba.stackexchange.com/questions/172230/howto-debug-slave-sql-error-… any idea? Lukas

1 0

[Maria-discuss] M|17 MariaDB Conference 2017 videos
by Lukas Lehner 27 Apr '17

27 Apr '17

Hi where can I watch M|17 video recordings? Is this planned? Will it be free or only for customers? Lukas

2 1

Re: [Maria-discuss] MySQL Workbench Fork?
by Federico Razzoli 27 Apr '17

27 Apr '17

I've just seen this Quora answer, which lists some good reasons to avoid GUIs. There are others. Again: my point is not "GUI's shouldn't exist", it's just "please don't call them an enterprise software, because they are not". https://www.quora.com/Do-database-administrators-actually-think-that-GUIs-a… Cheers, Federico

2 1

[Maria-discuss] userstats Denied_connections
by Rhys.Campbell＠swisscom.com 27 Apr '17

27 Apr '17

Hello All, I've been using the userstats plugin do debug an application issue and I think there's a bug in the collection of statistics, or at least a case of bad wording... https://mariadb.com/kb/en/mariadb/user-statistics/ DENIED_CONNECTIONS - The number of connections denied to this user. ACCESS_DENIED - The number of times this user's connections issued commands that were denied. A application user was displaying very high "DENIED_CONNECTION" values. This sent us in the direction a bad password was being supplied. Long story short it actually turns out that a series of ALTER TABLE statements were failing, because the application does not have permission to do so. During one our tests DENIED_CONNECTION was 5397 and ACCESS_DENIED was 0. We can see the ALTER TABLE statement failing and DENIED_CONNECTION being incremented quickly. According to the documentation shouldn't this ACCESS_DENIED (which I think might be better named COMMAND_DENIED?). Cheers, Rhys

2 1

[Maria-discuss] Inconsistent Query Results from Dependent Subquery
by Sam Vaughan 27 Apr '17

27 Apr '17

I'm seeing some odd behaviour on MariaDB 10.1.22 that's got me scratching my head. My production servers (a master and two slaves) are returning incorrect query results, but when I run the same query on the same data elsewhere I always get the correct results, even on a VM with matching OS, H/W arch and MariaDB version. The problem is triggered by using "ORDER BY x DESC LIMIT 1" in a dependent subquery on a single table. The query is being run on a small InnoDB table that very rarely changes. Each row contains a date, an integer identifier, a sequence number and a handful of applicable settings that are all integers. Unfortunately I can't create a simplified reproducible test case, but the following example shows the sort of query I'm working with. Imagine a table that looks like this: CREATE TABLE `settings` ( `date` date NOT NULL, `id` int(11) NOT NULL, `seq` int(11) NOT NULL, `value` int(11) NOT NULL, UNIQUE KEY `date_id_seq` (`date`,`id`,`seq`) ) ENGINE=InnoDB DEFAULT CHARSET=latin1; The table is designed to store a sequence of values belonging to something identified by the ID. Each sequence of values has a date stamp, and can be replaced at a later date by a new sequence of values: +------------+----+-----+-------+ | date | id | seq | value | +------------+----+-----+-------+ | 2011-11-01 | 1 | 0 | 6 | | 2011-11-01 | 1 | 1 | 3 | | 2011-11-01 | 1 | 2 | 9 | | 2011-11-01 | 2 | 0 | 6 | | 2011-11-01 | 2 | 1 | 1 | | 2011-11-01 | 3 | 0 | 6 | | 2011-11-01 | 3 | 1 | 8 | | 2012-04-01 | 1 | 0 | 4 | | 2012-04-01 | 1 | 1 | 3 | | 2013-06-01 | 3 | 0 | 5 | +------------+----+-----+-------+ The query is asking for all the IDs that had a particular first value set as of a particular date, in this case a first value of 6 as of '2012-07-01': SELECT s1.id FROM settings s1 WHERE s1.seq = 0 and s1.value = 6 and s1.date = ( SELECT s2.date FROM settings s2 WHERE s2.id = s1.id and s2.date <= '2012-07-01' ORDER BY s2.date DESC LIMIT 1 ); This query correctly returns IDs 2 and 3, and not ID 1 since the data for '2012-04-01' overrides the data for '2011-11-01'. In my equivalent situation, my production servers are returning 1, 2 and 3. Everywhere else I try to run it I get just 2 and 3 as expected. The table in question changes rarely, so it's trivial to show that the test system I've set up is working on the same data as the production systems. I've selected the entire table to out-files and diffed them, I've restored the test system from a mysqldump backup of production, and I've even run XtraBackup and prepared it with --export, then followed the instructions to import the table files into my test system just in case there's something weird in the production table's .ibd file. No matter what I do, I can't reproduce the problem anywhere but on my production systems. Testing has shown that minor changes to the query will fix the result set. Changing it to use "max()" instead of "order/limit" fixes it. Removing the date test from the where clause also fixes it. I don't want to just change all my queries like this to use MAX() and be done with it, because this weirdness will always be in the back of my mind. Is my query insufficiently specific? Have my queries of this type only ever returned the right results by accident until I moved to MariaDB recently? I haven't tried adding an index to the date column yet, because it too might make the problem vanish and I'll be none the wiser. I'd rather work out how to make the problem reproducible then dig in further to expose the cause. I thought that my InnoDB file could have been corrupt, but preparing the XtraBackup didn't detect it. Perhaps I've triggered a bug in MariaDB that needs very special conditions to appear? That seems unlikely but I'm at a bit of a loss now. Any suggestions anyone might have would be very gratefully received! Kind regards, Sam

1 0