FOSDEM 2022 is going to happen on the next weekend and I am still missing blog posts supporting my unusual upcoming talk there, this time devoted to building and using MariaDB server on macOS. So tonight I am going to contribute something new to refer to during my talk or followup questions. I'll try to document my way to build MariaDB Server from GitHub source on old (early 2015) MacBook Air running macOS 10.13.6 High Sierra. I am also going to show why one may want to use macOS there instead of installing Linux and having way more usual environment on the same decent hardware.
I've inherited this Air from my daughter in September, 2021 after she had got a new, M1-based, and initially used it mostly for Zoom calls and web-browsing. Soon I recalled that I've used MacBook for more than 3 years in the past while working for Sun and Oracle, and it was my main working platform not only for content consumption, emails and virtual machines. It was regularly used for bugs verification, MySQL builds and tests of all kinds. It would be a waste of capable OS and good hardware (still more powerful formally than all my other machines but Fedora 33 desktop) NOT to try to use it properly.
That's why after few software upgrades to end up with more recent 10.13 minor release:
Yuliyas-Air:maria10.6 Valerii$ uname -a
Darwin Yuliyas-Air 17.7.0 Darwin Kernel Version 17.7.0: Mon Aug 31 22:11:23 PDT 2020; root:xnu-4570.71.82.6~1/RELEASE_X86_64 x86_64
I read the manual, registered as a developer, downloaded and installed proper version of XCode and decided to proceed with MacPorts.
Then I updated ports tree and proceed by the manual, with installing git and other surely needed tools and packages:
sudo port install git cmake jemalloc judy openssl boost gnutls
That was just the beginning, and eventually, with dependencies, port updates, problems and workarounds, I ended up like this (and that works for building current code of 10.1 to 10.8 for sure, with some ports maybe not needed or used for other builds, but who cares):
Yuliyas-Air:maria10.8 Valerii$ port installed
The following ports are currently installed:
autoconf @2.71_1 (active)
automake @1.16.5_0 (active)
bison @3.8.2_0
bison @3.8.2_2 (active)
bison-runtime @3.8.2_0 (active)
boehmgc @8.0.6_0 (active)
boost @1.76_0 (active)
boost171 @1.71.0_3+no_single+no_static+python39 (active)
boost176 @1.76.0_2+no_single+no_static+python39
boost176 @1.76.0_3+no_single+no_static+python39 (active)
bzip2 @1.0.8_0 (active)
cmake @3.21.4_0
cmake @3.22.1_0 (active)
curl @7.80.0_0+ssl (active)
curl-ca-bundle @7.80.0_0
curl-ca-bundle @7.80.0_1 (active)
cyrus-sasl2 @2.1.27_5+kerberos (active)
db48 @4.8.30_4 (active)
expat @2.4.1_0
expat @2.4.2_0
expat @2.4.3_0 (active)
gdb @11.1_0 (active)
gdbm @1.22_0 (active)
gettext @0.21_0 (active)
gettext-runtime @0.21_0 (active)
gettext-tools-libs @0.21_0 (active)
git @2.34.1_1+credential_osxkeychain+diff_highlight+doc+pcre+perl5_28
git @2.34.1_2+credential_osxkeychain+diff_highlight+doc+pcre+perl5_28 (active)
gmp @6.2.1_0 (active)
gnutls @3.6.16_1
gnutls @3.6.16_2 (active)
icu @67.1_4 (active)
jemalloc @5.2.1_1 (active)
judy @1.0.5_1 (active)
kerberos5 @1.19.2_1 (active)
libarchive @3.5.2_1 (active)
libb2 @0.98.1_1 (active)
libcomerr @1.45.6_0 (active)
libcxx @5.0.1_4 (active)
libedit @20210910-3.1_1 (active)
libevent @2.1.12_1 (active)
libffi @3.4.2_2 (active)
libiconv @1.16_1 (active)
libidn @1.38_0 (active)
libidn2 @2.3.2_0
libidn2 @2.3.2_1 (active)
libpsl @0.21.1-20210726_2 (active)
libtasn1 @4.18.0_0 (active)
libtextstyle @0.21_0 (active)
libtool @2.4.6_13 (active)
libunistring @0.9.10_0
libunistring @1.0_0 (active)
libuv @1.42.0_1
libuv @1.43.0_0 (active)
libxml2 @2.9.12_1 (active)
libxslt @1.1.34_6 (active)
lmdb @0.9.29_0 (active)
luajit @2.1.0-beta3_5 (active)
lz4 @1.9.3_1 (active)
lzma @4.65_1 (active)
lzo2 @2.10_0 (active)
m4 @1.4.19_1 (active)
mariadb @5.5.68_0 (active)
mariadb-server @5.5.68_0 (active)
mysql57 @5.7.36_1 (active)
mysql_select @0.1.2_4 (active)
ncurses @6.3_0 (active)
nettle @3.7.3_0 (active)
openssl @3_1
openssl @3_2 (active)
openssl3 @3.0.0_6+legacy
openssl3 @3.0.1_0+legacy (active)
openssl11 @1.1.1l_5 (active)
p5-dbd-mysql @4.50.0_0 (active)
p5.28-authen-sasl @2.160.0_0 (active)
p5.28-cgi @4.530.0_0 (active)
p5.28-clone @0.450.0_0 (active)
p5.28-dbd-mysql @4.50.0_0+mysql57 (active)
p5.28-dbi @1.643.0_0 (active)
p5.28-digest-hmac @1.40.0_0 (active)
p5.28-digest-sha1 @2.130.0_4 (active)
p5.28-encode @3.160.0_0 (active)
p5.28-encode-locale @1.50.0_0 (active)
p5.28-error @0.170.290_0 (active)
p5.28-gssapi @0.280.0_3 (active)
p5.28-html-parser @3.760.0_0 (active)
p5.28-html-tagset @3.200.0_4 (active)
p5.28-http-date @6.50.0_0 (active)
p5.28-http-message @6.350.0_0 (active)
p5.28-io-html @1.4.0_0 (active)
p5.28-io-socket-ssl @2.72.0_0
p5.28-io-socket-ssl @2.73.0_0 (active)
p5.28-lwp-mediatypes @6.40.0_0 (active)
p5.28-mozilla-ca @20211001_0 (active)
p5.28-net-libidn @0.120.0_5 (active)
p5.28-net-smtp-ssl @1.40.0_0 (active)
p5.28-net-ssleay @1.900.0_4 (active)
p5.28-term-readkey @2.380.0_0 (active)
p5.28-time-local @1.300.0_0 (active)
p5.28-timedate @2.330.0_0 (active)
p5.28-uri @5.100.0_0 (active)
p11-kit @0.24.0_1 (active)
pcre2 @10.39_0 (active)
perl5.28 @5.28.3_4 (active)
perl5.30 @5.30.3_3 (active)
pkgconfig @0.29.2_0 (active)
popt @1.18_1 (active)
python3_select @0.0_2 (active)
python39 @3.9.9_0+lto+optimizations
python39 @3.9.10_0+lto+optimizations (active)
python_select @0.3_9 (active)
readline @8.1.000_0 (active)
readline-5 @5.2.014_2 (active)
rsync @3.2.3_1 (active)
sqlite3 @3.37.0_0
sqlite3 @3.37.1_0
sqlite3 @3.37.2_0 (active)
sysbench @1.0.20_0 (active)
tcp_wrappers @20_4 (active)
texinfo @6.8_0 (active)
umem @1.0.1_1 (active)
xxhashlib @0.8.1_0
xxhashlib @0.8.1_1 (active)
xz @5.2.5_0 (active)
zlib @1.2.11_0 (active)
zstd @1.5.0_0
zstd @1.5.1_0 (active)
Yuliyas-Air:maria10.8 Valerii$
This is surely not the minimal needed set of ports. I've highlighted a couple (like openssl11) that were really needed to build and install 10.8 successfully eventually.
Then I cloned the code with usual steps and ended up with this:
Yuliyas-Air:server Valerii$ git log -1
commit c1cef1afa9962544de4840c9a796ae0a9b5e92e6 (HEAD -> 10.8, origin/bb-10.8-wlad, origin/bb-10.8-release, origin/HEAD, origin/10.8)
Merge: db2013787d2 9d93b51effd
Author: Vladislav Vaintroub <wlad@mariadb.com>
Date: Wed Jan 26 13:57:00 2022 +0100
Merge remote-tracking branch 'origin/bb-10.8-wlad' into 10.8
Yuliyas-Air:server Valerii$ git submodule update --init --recursive
Submodule path 'libmariadb': checked out 'ddb031b6a1d8b6e26a0f10f454dc1453a48a6ca8'
Yuliyas-Air:server Valerii$ cd buildtmp/
Yuliyas-Air:buildtmp Valerii$ rm -rf *
Yuliyas-Air:buildtmp Valerii$ cmake .. -DCMAKE_INSTALL_PREFIX=/Users/Valerii/dbs/maria10.8 -DCMAKE_BUILD_TYPE=RelWithDebInfo -DBUILD_CONFIG=mysql_release -DFEATURE_SET=community -DWITH_EMBEDDED_SERVER=OFF -DPLUGIN_TOKUDB=NO -DWITH_SSL=/opt/local/libexec/openssl11 -DENABLE_DTRACE=1
...
-- The following OPTIONAL packages have not been found:
* PMEM
* Snappy
-- Configuring done
CMake Warning (dev):
Policy CMP0042 is not set: MACOSX_RPATH is enabled by default. Run "cmake
--help-policy CMP0042" for policy details. Use the cmake_policy command to
set the policy and suppress this warning.
MACOSX_RPATH is not specified for the following targets:
libmariadb
This warning is for project developers. Use -Wno-dev to suppress it.
-- Generating done
-- Build files have been written to: /Users/Valerii/git/server/buildtmp
Yuliyas-Air:buildtmp Valerii$
Note that I've used somewhat nontrivial cmake command line for out of source build, and the output above was not from the clean state, but from the state after 10.1 to 10.7 where all checked out and built successfully, one by one, with problems found and resolved (more on that on my slides and during the talk).
Two key options in that command line are -DWITH_SSL=/opt/local/libexec/openssl11 to use supported OpenSSL version 1.1 (even though for 10.8 default version 3 should already work too) and -DENABLE_DTRACE=1 to enable static DTrace probes in the code of MariaDB that are still there. The rest are typical for my builds and blog posts.
Then I proceeded with usual make to end up with the first problem:
Yuliyas-Air:buildtmp Valerii$ make
...
[ 46%] Built target heap
[ 46%] Building CXX object storage/innobase/CMakeFiles/innobase.dir/btr/btr0btr.cc.o
In file included from /Users/Valerii/git/server/storage/innobase/btr/btr0btr.cc:28:
In file included from /Users/Valerii/git/server/storage/innobase/include/btr0btr.h:31:
In file included from /Users/Valerii/git/server/storage/innobase/include/dict0dict.h:32:
In file included from /Users/Valerii/git/server/storage/innobase/include/dict0mem.h:45:
In file included from /Users/Valerii/git/server/storage/innobase/include/buf0buf.h:33:
/Users/Valerii/git/server/storage/innobase/include/fil0fil.h:1497:11: error:
'asm goto' constructs are not supported yet
__asm__ goto("lock btsl $31, %0\t\njnc %l1" : : "m" (n_pending)
^
1 error generated.
make[2]: *** [storage/innobase/CMakeFiles/innobase.dir/btr/btr0btr.cc.o] Error 1
make[1]: *** [storage/innobase/CMakeFiles/innobase.dir/all] Error 2
make: *** [all] Error 2
caused by the fact that clang 10.0.0 from XCode:
Yuliyas-Air:maria10.8 Valerii$ clang --version
Apple LLVM version 10.0.0 (clang-1000.10.44.4)
Target: x86_64-apple-darwin17.7.0
Thread model: posix
InstalledDir: /Library/Developer/CommandLineTools/usr/bin
does NOT support asm goto used in MariaDB Server code since 10.6 (while it was supposed to support it). I've applied a lame fix (see MDEV-27402 for more details and final diff later) and proceeded:
Yuliyas-Air:buildtmp Valerii$ make
...
[ 97%] Building CXX object client/CMakeFiles/mariadb.dir/mysql.cc.o
/Users/Valerii/git/server/client/mysql.cc:2853:59: error: expected expression
rl_attempted_completion_function= (rl_completion_func_t*)&new_mysql_co...
^
/Users/Valerii/git/server/client/mysql.cc:2853:38: error: use of undeclared
identifier 'rl_completion_func_t'; did you mean 'rl_completion_matches'?
rl_attempted_completion_function= (rl_completion_func_t*)&new_mysql_co...
^~~~~~~~~~~~~~~~~~~~
rl_completion_matches
/usr/include/editline/readline.h:202:16: note: 'rl_completion_matches' declared
here
char **rl_completion_matches(const char *, rl_compentry_func_t *);
^
/Users/Valerii/git/server/client/mysql.cc:2854:33: error: assigning to
'Function *' (aka 'int (*)(const char *, int)') from incompatible type
'rl_compentry_func_t *' (aka 'char *(*)(const char *, int)'): different
return type ('int' vs 'char *')
rl_completion_entry_function= (rl_compentry_func_t*)&no_completion;
^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
/Users/Valerii/git/server/client/mysql.cc:2856:3: error: no matching function
for call to 'rl_add_defun'
rl_add_defun("magic-space", (rl_command_func_t *)&fake_magic_space, -1);
^~~~~~~~~~~~
/usr/include/editline/readline.h:195:7: note: candidate function not viable: no
known conversion from 'rl_command_func_t *' (aka 'int (*)(int, int)') to
'Function *' (aka 'int (*)(const char *, int)') for 2nd argument
int rl_add_defun(const char *, Function *, int);
^
4 errors generated.
make[2]: *** [client/CMakeFiles/mariadb.dir/mysql.cc.o] Error 1
make[1]: *** [client/CMakeFiles/mariadb.dir/all] Error 2
make: *** [all] Error 2
Yuliyas-Air:buildtmp Valerii$
This readline-related problem was also reported (see MDEV-27579) and I fixed it with a lame patch, and eventually I was able to build successfully:
...
[100%] Linking C executable wsrep_check_version
[100%] Built target wsrep_check_version
Yuliyas-Air:buildtmp Valerii$ make install && make clean
...
-- Installing: /Users/Valerii/dbs/maria10.8/share/aclocal/mysql.m4
-- Installing: /Users/Valerii/dbs/maria10.8/support-files/mysql.server
Yuliyas-Air:buildtmp Valerii$ echo $?
0
The final diff is like this:
Yuliyas-Air:server Valerii$ git diff -u
diff --git a/client/mysql.cc b/client/mysql.cc
index 6612b273d17..902589f2e83 100644
--- a/client/mysql.cc
+++ b/client/mysql.cc
@@ -2849,7 +2849,7 @@ static void initialize_readline ()
rl_terminal_name= getenv("TERM");
/* Tell the completer that we want a crack first. */
-#if defined(USE_NEW_READLINE_INTERFACE)
+#if defined(USE_NEW_READLINE_INTERFACE) && !defined(__APPLE_CC__)
rl_attempted_completion_function= (rl_completion_func_t*)&new_mysql_completion;
rl_completion_entry_function= (rl_compentry_func_t*)&no_completion;
@@ -2859,7 +2859,9 @@ static void initialize_readline ()
setlocale(LC_ALL,""); /* so as libedit use isprint */
#endif
rl_attempted_completion_function= (CPPFunction*)&new_mysql_completion;
+#if !defined(__APPLE_CC__)
rl_completion_entry_function= &no_completion;
+#endif
rl_add_defun("magic-space", (Function*)&fake_magic_space, -1);
#else
rl_attempted_completion_function= (CPPFunction*)&new_mysql_completion;
diff --git a/storage/innobase/include/fil0fil.h b/storage/innobase/include/fil0fil.h
index 34a53746b42..a795313116f 100644
--- a/storage/innobase/include/fil0fil.h
+++ b/storage/innobase/include/fil0fil.h
@@ -1489,7 +1489,7 @@ inline void fil_space_t::reacquire()
inline bool fil_space_t::set_stopping_check()
{
mysql_mutex_assert_owner(&fil_system.mutex);
-#if defined __clang_major__ && __clang_major__ < 10
+#if (defined __clang_major__ && __clang_major__ < 10) || defined __APPLE_CC__
/* Only clang-10 introduced support for asm goto */
return n_pending.fetch_or(STOPPING, std::memory_order_relaxed) & STOPPING;
#elif defined __GNUC__ && (defined __i386__ || defined __x86_64__)
diff --git a/storage/rocksdb/rocksdb b/storage/rocksdb/rocksdb
--- a/storage/rocksdb/rocksdb
+++ b/storage/rocksdb/rocksdb
@@ -1 +1 @@
-Subproject commit bba5e7bc21093d7cfa765e1280a7c4fdcd284288
+Subproject commit bba5e7bc21093d7cfa765e1280a7c4fdcd284288-dirty
Yuliyas-Air:server Valerii$
The last diff in rocksdb submodule is related to the "missing zstd headers" problem I fined with yet another lame patch while building 10.7, see MDEV-27619.
After usual mysql_install_db and startup, I've ended up with a shiny new MariaDB 10.8.0 up and running on macOS 10.13.6:
Yuliyas-Air:maria10.8 Valerii$ bin/mysql
Welcome to the MariaDB monitor. Commands end with ; or \g.
Your MariaDB connection id is 5
Server version: 10.8.0-MariaDB MariaDB Server
Copyright (c) 2000, 2018, Oracle, MariaDB Corporation Ab and others.
Type 'help;' or '\h' for help. Type '\c' to clear the current input statement.
MariaDB [(none)]> show variables like 'version%';
+-------------------------+------------------------------------------+
| Variable_name | Value |
+-------------------------+------------------------------------------+
| version | 10.8.0-MariaDB |
| version_comment | MariaDB Server |
| version_compile_machine | x86_64 |
| version_compile_os | osx10.13 |
| version_malloc_library | system |
| version_source_revision | c1cef1afa9962544de4840c9a796ae0a9b5e92e6 |
| version_ssl_library | OpenSSL 1.1.1l 24 Aug 2021 |
+-------------------------+------------------------------------------+
7 rows in set (0.001 sec)
I even checked few MTR test suites, and many tests pass:
Yuliyas-Air:mysql-test Valerii$ ./mtr --suite=rocksdb
Logging: ./mtr --suite=rocksdb
VS config:
vardir: /Users/Valerii/dbs/maria10.8/mysql-test/var
Checking leftover processes...
- found old pid 89161 in 'mysqld.1.pid', killing it...
process did not exist!
Removing old var directory...
Creating var directory '/Users/Valerii/dbs/maria10.8/mysql-test/var'...
Checking supported features...
MariaDB Version 10.8.0-MariaDB
- SSL connections supported
Using suites: rocksdb
Collecting tests...
Installing system database...
...
rocksdb.index_merge_rocksdb 'write_committed' [ pass ] 911
rocksdb.shutdown 'write_prepared' [ pass ] 1895
rocksdb.index_merge_rocksdb 'write_prepared' [ pass ] 923
rocksdb.mariadb_misc_binlog 'write_committed' [ pass ] 43
rocksdb.mariadb_misc_binlog 'write_prepared' [ pass ] 43
...
rocksdb.issue495 'write_committed' [ pass ] 33
rocksdb.partition 'write_committed' [ fail ]
Test ended at 2022-01-27 22:41:15
CURRENT_TEST: rocksdb.partition
mysqltest: At line 67: query 'ALTER TABLE t1 REBUILD PARTITION p0, p1' failed: ER_METADATA_INCONSISTENCY (4064): Table 'test.t1#P#p0#TMP#' does not exist, but metadata information exists inside MyRocks. This is a sign of data inconsistency. Please check if './test/t1#P#p0#TMP#.frm' exists, and try to restore it if it does not exist.
The result from queries just before the failure was:
< snip >
DROP TABLE IF EXISTS employees_hash_1;
DROP TABLE IF EXISTS t1_hash;
DROP TABLE IF EXISTS employees_linear_hash;
DROP TABLE IF EXISTS t1_linear_hash;
DROP TABLE IF EXISTS k1;
DROP TABLE IF EXISTS k2;
DROP TABLE IF EXISTS tm1;
DROP TABLE IF EXISTS tk;
DROP TABLE IF EXISTS ts;
DROP TABLE IF EXISTS ts_1;
DROP TABLE IF EXISTS ts_3;
DROP TABLE IF EXISTS ts_4;
DROP TABLE IF EXISTS ts_5;
DROP TABLE IF EXISTS trb3;
DROP TABLE IF EXISTS tr;
DROP TABLE IF EXISTS members_3;
DROP TABLE IF EXISTS clients;
DROP TABLE IF EXISTS clients_lk;
DROP TABLE IF EXISTS trb1;
CREATE TABLE t1 (i INT, j INT, k INT, PRIMARY KEY (i)) ENGINE = ROCKSDB PARTITION BY KEY(i) PARTITIONS 4;
More results from queries before failure can be found in /Users/Valerii/dbs/maria10.8/mysql-test/var/log/partition.log
- saving '/Users/Valerii/dbs/maria10.8/mysql-test/var/log/rocksdb.partition-write_committed/' to '/Users/Valerii/dbs/maria10.8/mysql-test/var/log/rocksdb.partition-write_committed/'
Only 148 of 555 completed.
--------------------------------------------------------------------------
The servers were restarted 23 times
Spent 154.726 of 206 seconds executing testcases
Failure: Failed 1/27 tests, 96.30% were successful.
Failing test(s): rocksdb.partition
The log files in var/log may give you some hint of what went wrong.
If you want to report this error, please read first the documentation
at http://dev.mysql.com/doc/mysql/en/mysql-test-suite.html
69 tests were skipped, 2 by the test itself.
mysql-test-run: *** ERROR: there were failing test cases
But the real reason to build MariaDB server on macOS (other than "because I can") was not even RocksDB testing, but this:
Yuliyas-Air:mysql-test Valerii$ dtrace
Usage: dtrace [-aACeFHlqSvVwZ] [-arch i386|x86_64] [-b bufsz] [-c cmd] [-D name[=def]]
[-I path] [-L path] [-o output] [-p pid] [-s script] [-U name]
[-x opt[=val]]
[-P provider [[ predicate ] action ]]
[-m [ provider: ] module [[ predicate ] action ]]
[-f [[ provider: ] module: ] func [[ predicate ] action ]]
[-n [[[ provider: ] module: ] func: ] name [[ predicate ] action ]]
[-i probe-id [[ predicate ] action ]] [ args ... ]
predicate -> '/' D-expression '/'
action -> '{' D-statements '}'
-arch Generate programs and Mach-O files for the specified architecture
-a claim anonymous tracing state
-A generate plist(5) entries for anonymous tracing
-b set trace buffer size
-c run specified command and exit upon its completion
-C run cpp(1) preprocessor on script files
-D define symbol when invoking preprocessor
-e exit after compiling request but prior to enabling probes
-f enable or list probes matching the specified function name
-F coalesce trace output by function
-h generate a header file with definitions for static probes
-H print included files when invoking preprocessor
-i enable or list probes matching the specified probe id
-I add include directory to preprocessor search path
-l list probes matching specified criteria
-L add library directory to library search path
-m enable or list probes matching the specified module name
-n enable or list probes matching the specified probe name
-o set output file
-p grab specified process-ID and cache its symbol tables
-P enable or list probes matching the specified provider name
-q set quiet mode (only output explicitly traced data)
-s enable or list probes according to the specified D script
-S print D compiler intermediate code
-U undefine symbol when invoking preprocessor
-v set verbose mode (report stability attributes, arguments)
-V report DTrace API version
-w permit destructive actions
-W wait for specified process and exit upon its completion
-x enable or modify compiler and tracing options
-Z permit probe descriptions that match zero probes
Yuliyas-Air:mysql-test Valerii$
I remember how cool dtrace was since workin for Sun and after last re-install of Fedora I ended up with my Illumos VM gone and no recent FreeBSD VM anyway,. so basically there was no dtrace at hand until I've got this MacBook.
So, unlike MySQL 8.0, MariaDB server still contains USDT (DTrace probes) and I've built my version with them enabled. Let's quickly check how they can be used. There are few sample D files in the source:
Yuliyas-Air:server Valerii$ ls support-files/dtrace/
locktime.d query-rowops.d
query-execandqc.d query-time.d
query-filesort-time.d statement-time.d
query-network-time.d statement-type-aggregate.d
query-parse-time.d
Yuliyas-Air:server Valerii$ cat support-files/dtrace/query-time.d
#!/usr/sbin/dtrace -s
#
# Copyright (c) 2009 Sun Microsystems, Inc.
# Use is subject to license terms.
#
# This program is free software; you can redistribute it and/or modify
# it under the terms of the GNU General Public License as published by
# the Free Software Foundation; version 2 of the License.
#
# This program is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
# GNU General Public License for more details.
#
# You should have received a copy of the GNU General Public License
# along with this program; if not, write to the Free Software
# Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1335 USA
#
# Shows basic query execution time, who execute the query, and on what database
#pragma D option quiet
dtrace:::BEGIN
{
printf("%-20s %-20s %-40s %-9s\n", "Who", "Database", "Query", "Time(ms)");
}
mysql*:::query-start
{
self->query = copyinstr(arg0);
self->connid = arg1;
self->db = copyinstr(arg2);
self->who = strjoin(copyinstr(arg3),strjoin("@",copyinstr(arg4)));
self->querystart = timestamp;
}
mysql*:::query-done
{
printf("%-20s %-20s %-40s %-9d\n",self->who,self->db,self->query,
(timestamp - self->querystart) / 1000000);
}
Yuliyas-Air:server Valerii$
We can try to run it (with yet another hack to do that I leave to the curious reader who is really going to try this) and then execute some queries against MariaDB from another terminal:
Yuliyas-Air:server Valerii$ sudo support-files/dtrace/query-time.d
dtrace: system integrity protection is on, some features will not be available
Who Database Query Time(ms)
Valerii@localhost select @@version_comment limit 1 0
Valerii@localhost select 1+1 0
Valerii@localhost select sleep(4) 4004
Valerii@localhost select @@version_comment limit 1 0
Valerii@localhost create database sbtest 1
Valerii@localhost sbtest CREATE TABLE sbtest1(
id INTEGER NOT NULL AUTO_INCREMENT,
k INTEGER DEFAULT '0' NOT NULL,
c CHAR(120) DEFAULT '' NOT NULL,
pad CHAR(60) DEFAULT '' NOT NULL,
PRIMARY KEY (id)
) /*! ENGINE = innodb */ 91
Valerii@localhost sbtest INSERT INTO sbtest1(k, c, pad) VALUES(366941, '31451373586-15688153734-79729593694-96509299839-83724898275-86711833539-78981337422-35049690573-51724173961-87474696253', '98996621624-36689827414-04092488557-09587706818-65008859162'),(277750, '21472970079-7 181
Valerii@localhost sbtest INSERT INTO sbtest1(k, c, pad) VALUES(124021, '80697810288-90543941719-80227288793-55278810422-59841440561-49369413842-83550451066-12907725305-62036548401-86959403176', '65708342793-83311865079-53224065384-18645733125-16333693298'),(660496, '07381386584-5 31
Valerii@localhost sbtest INSERT INTO sbtest1(k, c, pad) VALUES(425080, '48883413333-17783399741-03981526516-97596354402-27141206678-83563692683-30244461835-25263435890-49140039573-28211133426', '81560227417-96691828090-72817141653-15106797886-43970285630'),(322421, '68618246702-7 33
...
to end up with a lightweight general query log with query execution time repported in milliseconds. Note some statement from, yes, sysbench test preapre :)
 |
| Nice view. macOS is also nice and useful platform to run MariaDB server for fun :) |
Thta's probably enough for a simgle post. Now go and get that from your MySQL 8 on any OS! USDTs rule!
To summarize:
- It is surely possible to build even latest ang greates still alpha MariaDB 10.8 on macOS even as old as 10.13.6, and get usable result.
- One of the main benefits of macOS is DTrace support.
- It is still possible to use a limited set of USDTs once added to MySQL and now gone in all current versions of MariaDB server.
- That rocksdb.partition test failure is to be studied an maybe reported as a MDEV
- DTrace is surely more capable than wehat those examples in the source code show. More DTrace-related posts are coming this year. Stay tuned!
No comments:
Post a Comment