polardbxengine/mysql-test/suite/ndb/t/ndbinfo.test

1026 lines
32 KiB
Plaintext

--result_format 2
--source include/have_ndb.inc
SELECT PLUGIN_NAME,PLUGIN_VERSION,PLUGIN_STATUS,PLUGIN_TYPE,
PLUGIN_LIBRARY,PLUGIN_LIBRARY_VERSION,PLUGIN_AUTHOR,PLUGIN_DESCRIPTION
FROM information_schema.plugins WHERE PLUGIN_NAME = 'ndbinfo';
## Creation of temporary tables should not be supported by NDBINFO engine
--error ER_ILLEGAL_HA_CREATE_OPTION
CREATE TEMPORARY TABLE `t1` (
`dummy` INT UNSIGNED
) ENGINE=NDBINFO;
USE ndbinfo;
# Current usage may vary. Therefor the test only checks that if falls in the
# <0%,5%> range.
SELECT node_id, memory_type, used>0 AND used < total/20 AS used_ok,
used_pages>0 AND used_pages < total_pages/20 AS used_pages_ok,
total, total_pages
FROM ndbinfo.memoryusage WHERE memory_type="Long message buffer"
ORDER BY node_id;
# Current usage may vary. Therefor the test only checks that if falls in the
# <0%,5%> range.
SELECT node_id, block_number, block_instance, pool_name,
used>0 AND used < total/20 AS used_ok,
total, high>=used AND high < total/20 AS high_ok, entry_size, config_param1,
config_param2, config_param3, config_param4
FROM ndbinfo.ndb$pools WHERE pool_name="Long message buffer"
ORDER BY node_id;
SHOW CREATE TABLE ndb$tables;
SELECT * FROM ndb$tables;
SELECT COUNT(*) FROM ndb$tables;
SELECT * FROM ndb$tables WHERE table_id = 2;
SELECT * FROM ndb$tables WHERE table_id > 5;
SELECT * FROM ndb$tables WHERE table_name = 'LOGDESTINATION';
# Show a simple join is working
SELECT COUNT(*) FROM ndb$tables t, ndb$columns c
WHERE t.table_id = c.table_id AND
t.table_id in (1,2,3,4,5,6);
SELECT table_id, table_name, comment from ndb$tables
WHERE table_id > 2 AND table_id <= 5 ORDER BY table_id;
SELECT table_id FROM ndb$tables WHERE table_id = 2 ORDER BY table_name;
SELECT table_id, table_name FROM ndb$tables ORDER BY table_name;
SELECT table_id, column_id, column_name FROM ndb$columns LIMIT 7;
--error ER_OPEN_AS_READONLY
UPDATE ndb$tables SET table_id=2 WHERE table_id=3;
--error ER_OPEN_AS_READONLY
UPDATE ndb$tables SET table_id=9 WHERE 1=0;
--error ER_OPEN_AS_READONLY
UPDATE ndb$tables SET table_id=9 WHERE table_id > 1;
--error ER_OPEN_AS_READONLY
DELETE FROM ndb$tables WHERE table_id=3;
--error ER_OPEN_AS_READONLY
DELETE FROM ndb$tables WHERE 1=0;
--error ER_OPEN_AS_READONLY
DELETE FROM ndb$tables WHERE table_id > 1;
--error ER_ILLEGAL_HA
ALTER TABLE ndb$test ADD COLUMN another_col varchar(255);
FLUSH TABLES;
SELECT table_id FROM ndb$tables;
--error ER_OPEN_AS_READONLY
TRUNCATE ndb$tables;
## Variables and status
SHOW GLOBAL STATUS LIKE 'ndbinfo\_%';
let $current_version = `select @@ndbinfo_version`;
--replace_result $current_version NDB_VERSION_D
SHOW GLOBAL VARIABLES LIKE 'ndbinfo\_%';
SELECT counter, HEX(counter2) FROM ndb$test LIMIT 10;
# All tables that contain data are hidden by default
# and becomes visible with ndbinfo_show_hidden
SHOW TABLES LIKE 'ndb$te%';
set @@ndbinfo_show_hidden=TRUE;
SHOW TABLES LIKE 'ndb$te%';
set @@ndbinfo_show_hidden=default;
# Check that ndbinfo_table_prefix is readonly
--error ER_INCORRECT_GLOBAL_LOCAL_VAR
set @@ndbinfo_table_prefix="somethingelse";
# Check that ndbinfo_database is readonly
--error ER_INCORRECT_GLOBAL_LOCAL_VAR
set @@ndbinfo_database="somethingelse";
# Check that block table has been created and contain data
SELECT count(*) >= 20 FROM blocks;
# Test incompatible table definition between NDB and MySQL Server
# using the ndb$test table which originally looks like
show create table ndb$test;
# CREATE TABLE `ndb$test` (
# `node_id` int unsigned DEFAULT NULL,
# `block_number` int unsigned DEFAULT NULL,
# `block_instance` int unsigned DEFAULT NULL,
# `counter` int unsigned DEFAULT NULL,
# `counter2` bigint unsigned DEFAULT NULL
# ) ENGINE=NDBINFO COMMENT='for testing'
## 1) More columns in NDB -> allowed, with warning
DROP TABLE ndb$test;
CREATE TABLE ndb$test (node_id int unsigned) ENGINE = ndbinfo;
SELECT node_id != 0 FROM ndb$test LIMIT 1;
DROP TABLE ndb$test;
## 2) Column does not exist in NDB -> allowed, with warning, non existing
## column(s) return NULL
## 2a) Extra column at end
CREATE TABLE ndb$test (node_id int, non_existing int) ENGINE = ndbinfo;
SELECT DISTINCT node_id, non_existing FROM ndb$test;
DROP TABLE ndb$test;
## 2b) Extra column(s) in middle
CREATE TABLE ndb$test (
node_id int unsigned,
non_existing int unsigned,
block_number int unsigned,
block_instance int unsigned,
counter int unsigned,
counter2 bigint unsigned
) ENGINE = ndbinfo;
SELECT DISTINCT node_id, non_existing, block_number FROM ndb$test;
DROP TABLE ndb$test;
## 2c) Extra column first
CREATE TABLE ndb$test (non_existing int, node_id int) ENGINE = ndbinfo;
SELECT DISTINCT node_id, non_existing FROM ndb$test;
SELECT DISTINCT non_existing, node_id FROM ndb$test;
DROP TABLE ndb$test;
## 3) Incompatible column type -> error, with warning
## 3a) int instead of bigint
CREATE TABLE ndb$test (counter2 int) ENGINE = ndbinfo;
--error ER_GET_ERRMSG
SELECT * FROM ndb$test;
SHOW WARNINGS;
## 3b) bigint instead of int
DROP TABLE ndb$test;
CREATE TABLE ndb$test (node_id bigint) ENGINE = ndbinfo;
--error ER_GET_ERRMSG
SELECT * FROM ndb$test;
SHOW WARNINGS;
## 3c) varchar instead of int
DROP TABLE ndb$test;
CREATE TABLE ndb$test (node_id varchar(255)) ENGINE = ndbinfo;
--error ER_GET_ERRMSG
SELECT * FROM ndb$test;
SHOW WARNINGS;
DROP TABLE ndb$test;
## 3d) column which is NOT NULL
CREATE TABLE ndb$test (node_id int unsigned NOT NULL) ENGINE = ndbinfo;
--error ER_GET_ERRMSG
SELECT * FROM ndb$test;
SHOW WARNINGS;
DROP TABLE ndb$test;
## 3e) non existing column which is NOT NULL
CREATE TABLE ndb$test (
block_number int unsigned,
non_existing int NOT NULL) ENGINE = ndbinfo;
--error ER_GET_ERRMSG
SELECT * FROM ndb$test;
SHOW WARNINGS;
DROP TABLE ndb$test;
## 4) Table with primary key/indexes not supported
--error ER_TOO_MANY_KEYS
CREATE TABLE ndb$test (node_id int, block_number int PRIMARY KEY) ENGINE = ndbinfo;
## 5) Table with blobs not supported
--error ER_TABLE_CANT_HANDLE_BLOB
CREATE TABLE ndb$test (node_id int, block_number blob) ENGINE = ndbinfo;
## 6) Table with autoincrement not supported
--error ER_TABLE_CANT_HANDLE_AUTO_INCREMENT
CREATE TABLE ndb$test (node_id int AUTO_INCREMENT) ENGINE = ndbinfo;
## Restore original ndb$test table to avoid test side-effects
CREATE TABLE ndb$test (
node_id int(10) unsigned DEFAULT NULL,
block_number int(10) unsigned DEFAULT NULL,
block_instance int(10) unsigned DEFAULT NULL,
counter int(10) unsigned DEFAULT NULL,
counter2 bigint(20) unsigned DEFAULT NULL
) ENGINE=NDBINFO DEFAULT CHARSET=latin1 COMMENT='for testing';
# wl#5567 - exercise table...
# only node_id is guranteed to be same...
# (unless we change cluster config for test)
# but this will anyway pull results from datanode...
#
--sorted_result
select distinct node_id
from ndbinfo.diskpagebuffer;
#
# BUG#11885602
# - It was allowed to CREATE TABLE which was not in NDB, but
# creating a view on that table failed. Implement ndbinfo_offline
# mode which allows tables to be created and opened although they
# don't exists or have different table definition.
# This is exactly the same behaviour as when NDBCLUSTER
# is disabled
#
# Check ndbinfo_offline is GLOBAL variable
--error ER_GLOBAL_VARIABLE
set @@ndbinfo_offline=1;
# Query used to check that open tables are closed
# when offline mode is turned on and off
let $q1 = SELECT DISTINCT(node_id) FROM ndbinfo.counters ORDER BY node_id;
eval $q1;
# new views
desc threadblocks;
select distinct block_name from threadblocks order by 1;
desc threadstat;
select count(*) > 0 block_name from threadstat;
desc disk_write_speed_aggregate;
select count(*) from disk_write_speed_aggregate;
desc disk_write_speed_base;
select node_id from disk_write_speed_base group by node_id, thr_no;
select count(*) from disk_write_speed_aggregate_node;
desc restart_info;
select * from restart_info where node_restart_status <> 'Restart completed';
desc tc_time_track_stats;
select distinct node_id from tc_time_track_stats;
desc threads;
desc cluster_transactions;
desc server_transactions;
desc cluster_operations;
desc server_operations;
select count(*) > 0 from table_distribution_status;
select count(*) > 0 from ndb$table_distribution_status_all;
select count(*) > 0 from table_fragments;
select count(*) > 0 from ndb$table_fragments_all;
select count(*) > 0 from table_replicas;
select count(*) > 0 from ndb$table_replicas_all;
# DESCRIBE columns of the views
desc table_distribution_status;
desc table_fragments;
desc table_replicas;
# DESCRIBE columns of the hidden table
# (turn show_hidden on since otherwise columns
# of the table is supposed to be hidden)
set @@ndbinfo_show_hidden=TRUE;
desc ndb$table_distribution_status_all;
desc ndb$table_fragments_all;
desc ndb$table_replicas_all;
set @@ndbinfo_show_hidden=default;
create table t1 (a int primary key) engine = ndb;
begin;
insert into t1 values (1);
select state, count_operations, outstanding_operations,
IF(client_node_id <= 255, "<client_node_id>", "<incorrect node id>")
client_node_id
from server_transactions;
select node_id, operation_type, state,
IF(tc_node_id <= 48, "<tc_node_id>", "<incorrect nodeid>") tc_node_id,
IF(client_node_id <= 255, "<client_node_id>", "<incorrect node id>")
client_node_id
from server_operations
order by 1;
--sorted_result
select st.state, st.count_operations, st.outstanding_operations,
so.node_id, so.state, so.operation_type
from server_transactions st,
server_operations so
where st.transid = so.transid
and so.tc_node_id = st.node_id
and so.tc_block_instance = st.block_instance
and so.client_node_id = st.client_node_id
and so.client_block_ref = st.client_block_ref;
rollback;
drop table t1;
#
select distinct resource_name
from resources
order by 1;
select distinct counter_name
from counters
order by 1;
# Turn on ndbinfo_offline
set @@global.ndbinfo_offline=TRUE;
select @@ndbinfo_offline;
CREATE TABLE ndb$does_not_exist_in_ndb(
node_id int,
message varchar(255)
) ENGINE = ndbinfo CHARACTER SET latin1;
CREATE VIEW view_on_table_which_does_not_exist_in_ndb AS
SELECT node_id, message
FROM ndbinfo.ndb$does_not_exist_in_ndb;
SHOW CREATE TABLE ndb$does_not_exist_in_ndb;
# SELECTs return no rows in offline mode
SELECT * FROM view_on_table_which_does_not_exist_in_ndb;
SELECT * FROM ndb$does_not_exist_in_ndb;
eval $q1;
DROP VIEW view_on_table_which_does_not_exist_in_ndb;
DROP TABLE ndb$does_not_exist_in_ndb;
# Restore original value
set @@global.ndbinfo_offline = FALSE;
eval $q1;
#
# Bug#17018500 BOGUS DEFINER VALUE FOR NDBINFO VIEWS REPORTED
# IN INFORMATION_SCHEMA.VIEW
#
--echo Verify there are no ndbinfo views without 'root@localhost' as definer
select table_schema, table_name, definer from information_schema.views
where table_schema = 'ndbinfo' and
definer != 'root@localhost';
#
# Test buffered joining
#
# BUG#20075747 RND_INIT() ON AN OPEN SCAN IS USED TO REPOSITION THE CURSOR
# - rnd_init() on an open scan is used to reposition the "cursor"
# back to first row
#
set @@optimizer_switch='block_nested_loop=off';
--disable_result_log
select @a:=0;
select
@a:=@a+1,
a.node_id, b.node_id,
a.block_name, b.block_name,
a.block_instance, b.block_instance,
a.counter_id, b.counter_id,
a.counter_name, b.counter_name,
a.val, b.val
from
ndbinfo.counters a join
ndbinfo.counters b
on
a.node_id = b.node_id;
select @b:=0;
select
@b:=@b+1,
a.node_id, b.node_id,
a.block_name, b.block_name,
a.block_instance, b.block_instance,
a.counter_id, b.counter_id,
a.counter_name, b.counter_name,
a.val, b.val
from
ndbinfo.counters a join
ndbinfo.counters b
on
a.node_id = b.node_id;
--enable_result_log
--echo Expect both joins to give same row count
--echo (0 diff)
select @a - @b;
create temporary table old_count
select counter_name, sum(val) as val
from ndbinfo.counters
where block_name='DBSPJ'
group by counter_name;
drop table old_count;
set @@optimizer_switch=default;
# ################################
# Test memory_per_fragment table.
# ################################
# A helper table for populating t1.
create table numbers(
i int
);
insert into numbers values (0),(1),(2),(3),(4),(5),(6),(7),(8),(9);
create table t1(
a int not null primary key,
b int not null,
c1 char(255),
vc1 varchar(1024),
tx1 text
)engine=ndbcluster character set latin1;
# Popultate t1 with 1000 rows, including som sizable varsize data.
insert into t1 select x1.i+x2.i*10+x3.i*100, x1.i-x2.i*10-x3.i*100, repeat('x', 1+x1.i), repeat('y', 1000+x1.i), repeat('z', x1.i*100) from numbers as x1, numbers as x2, numbers as x3;
drop table numbers;
# Create some extra indexes to test different schema object types.
create index ix1 on t1(c1);
create unique index ix2 on t1(b) using hash;
# The table id of 't1' may vary from run to run. We need to make sure that the
# output from the subsequent tests does not depend on the numrical value of
# that table id.
set @t1_id = (select cast(id as signed integer) from ndbinfo.ndb$dict_obj_info where fq_name REGEXP "def/t1$");
# Test the dict_obj_info table by showing t1 and it's four releated objects.
# Replace table id of t1 in 'fq_name' values that contain it (index and blob
# table names do), since the table id may vary between runs.
select type, state, parent_obj_type,
replace(replace(fq_name,
concat("/", cast(@t1_id as char(10)), "/"), "/<t1 id>/"),
concat("_", cast(@t1_id as char(10)), "_"), "_<t1 id>_") as fq_name,
if(parent_obj_id=@t1_id, "child_of_t1",
if(parent_obj_id=0, "root", "error!")) as nesting
from ndbinfo.ndb$dict_obj_info
where type <= 6
and (id = @t1_id or
instr(fq_name, concat("/", cast(@t1_id as char(10)), "/")) or
instr(fq_name, concat("_", cast(@t1_id as char(10)), "_")))
order by type, state, parent_obj_type, fq_name;
# Disable query log while defining the view below, since view definition
# conatins the table id of t1, which may vary between runs.
disable_query_log;
let $t1_id = query_get_value(select @t1_id as id, id, 1);
# Create som views that we reuse after modifying t1 below.
# This one sums usage counts for all fragments of each table.
# Some values, such as fixed_elem_free_count and fixed_elem_count for ordered
# indexes may vary depending on the number of fragments and other factors.
# These are therefore not returned directly, but used in calculations that
# should provide stable results.
# Observe that memory_per_fragment.fq_name, parent_fq_name, type are all
# functionally dependent on 'table_id', but since this cannot be deduced from
# the schema, they are all included in the 'group by' clause.
# Exclude system tables and mysql schema since those may have different
# contents depending on earlier test runs.
eval create view sum_per_tab as
select replace(fq_name, cast($t1_id as char(10)), "<t1 id>") as fq_name,
parent_fq_name, type,
if (type="Ordered index",
-(round(sum(fixed_elem_count)/165)),
sum(fixed_elem_count)) as fixed_elem_count,
max(fixed_elem_size_bytes) as fixed_elem_size_bytes,
round(sum(fixed_elem_alloc_bytes/fixed_elem_size_bytes)
/sum(fixed_elem_count+fixed_elem_free_count),1) as fixed_elem_alloc_ok,
sum(fixed_elem_free_bytes/fixed_elem_size_bytes) =
sum(fixed_elem_free_count) as fixed_elem_free_bytes_ok,
sum(var_elem_count) as var_elem_count,
sum(var_elem_alloc_bytes) >= sum(var_elem_free_bytes) as var_bytes_ok
from ndbinfo.memory_per_fragment mpf
where not fq_name like "mysql/%" and
(parent_fq_name is null or not parent_fq_name like "mysql/%") and
type <> 'System table'
group by table_id, mpf.fq_name, parent_fq_name, type
order by ifnull(parent_fq_name, mpf.fq_name), mpf.fq_name, type;
enable_query_log;
# Sum usage counts for all fragments of all tables.
create view frag_sum as
select
sum(fixed_elem_alloc_bytes) as fixed_elem_alloc_bytes,
sum(var_elem_alloc_bytes) as var_elem_alloc_bytes,
sum(hash_index_alloc_bytes) as hash_index_alloc_bytes
from ndbinfo.memory_per_fragment;
# Sum pool values across all block instances.
# 'Data memory' appears twice in 'pools' therefore a 'where' clause is needed.
create view pool_sum as
select pool_name, block_number, max(entry_size) * sum(used) as used_bytes
from ndbinfo.ndb$pools where pool_name <> "Data memory"
or block_number=249 group by pool_name, block_number;
# Check that the combined size of all L2PMap nodes equals the total size
# of the allocated L2PMap pages minus a small (<1%) page overhead.
create view check_l2pmap_pools as
select min(floor(1000*nodes.total*nodes.entry_size/
(pages.used*pages.entry_size))) as l2pmap_ratio
from ndbinfo.ndb$pools as pages join ndbinfo.ndb$pools as nodes
where pages.node_id=nodes.node_id
and pages.block_instance=nodes.block_instance
and pages.block_number=nodes.block_number
and pages.pool_name="L2PMap pages"
and nodes.pool_name="L2PMap nodes";
create view l2p_frag_sum as
select sum(tuple_l2pmap_alloc_bytes) as l2p_sum, 'TUP' as block
from ndbinfo.ndb$frag_mem_use
union
select sum(hash_index_l2pmap_alloc_bytes) as l2p_sum, 'ACC' as block
from ndbinfo.ndb$frag_mem_use;
# Dump some extra info for solving Bug #28892753
let $DUMP_INFO=0;
# Check that we get the same result from the common pool as when we add up
# the values for each fragment.
select used_bytes - hash_index_alloc_bytes from pool_sum, frag_sum
where pool_name='Index memory';
select used_bytes - fixed_elem_alloc_bytes - var_elem_alloc_bytes
into @diff
from pool_sum, frag_sum
where pool_name='Data memory';
select @diff;
let $DIFF=query_get_value("select @diff as d", d, 1);
if ($DIFF)
{
let $DUMP_INFO=1;
}
select l2p_sum-used_bytes from l2p_frag_sum, pool_sum
where pool_name='L2PMap nodes' and ((block='ACC' and block_number=248) or
(block='TUP' and block_number=249));
select * from check_l2pmap_pools;
select * from sum_per_tab;
if ($DUMP_INFO)
{
# Extra debugging info
select * from sum_per_tab;
select * from frag_sum;
select * from pool_sum;
select * from check_l2pmap_pools;
select * from l2p_frag_sum;
select * from ndbinfo.ndb$pools;
select * from ndbinfo.resources;
}
# Now delete som scattered rows and check memory usage again.
delete from t1 where a%10=0;
select used_bytes - hash_index_alloc_bytes from pool_sum, frag_sum
where pool_name='Index memory';
select used_bytes - fixed_elem_alloc_bytes - var_elem_alloc_bytes
from pool_sum, frag_sum
where pool_name='Data memory';
select l2p_sum-used_bytes from l2p_frag_sum, pool_sum
where pool_name='L2PMap nodes' and ((block='ACC' and block_number=248) or
(block='TUP' and block_number=249));
select * from check_l2pmap_pools;
select * from sum_per_tab;
if ($DUMP_INFO)
{
# Extra debugging info
select * from sum_per_tab;
select * from frag_sum;
select * from pool_sum;
select * from check_l2pmap_pools;
select * from l2p_frag_sum;
select * from ndbinfo.ndb$pools;
select * from ndbinfo.resources;
}
# Now update some var-sized fields and check memory usage again.
update t1 set vc1=repeat('x', a%300), tx1=repeat('y', a*2);
select used_bytes - hash_index_alloc_bytes from pool_sum, frag_sum
where pool_name='Index memory';
select used_bytes - fixed_elem_alloc_bytes - var_elem_alloc_bytes
from pool_sum, frag_sum
where pool_name='Data memory';
select l2p_sum-used_bytes from l2p_frag_sum, pool_sum
where pool_name='L2PMap nodes' and ((block='ACC' and block_number=248) or
(block='TUP' and block_number=249));
select * from check_l2pmap_pools;
select * from sum_per_tab;
if ($DUMP_INFO)
{
# Extra debugging info
select * from sum_per_tab;
select * from frag_sum;
select * from pool_sum;
select * from check_l2pmap_pools;
select * from l2p_frag_sum;
select * from ndbinfo.ndb$pools;
select * from ndbinfo.resources;
}
drop table t1;
drop view l2p_frag_sum;
drop view check_l2pmap_pools;
drop view sum_per_tab;
drop view pool_sum;
drop view frag_sum;
# ################################
# Test operations_per_fragment table.
# ################################
# Create a view that sums counters across all fragments, so that results
# will not depend on endianness.
# Observe that operations_per_fragment.fq_name, parent_fq_name, type are all
# functionally dependent on 'table_id', but since this cannot be deduced from
# the schema, they are all included in the 'group by' clause.
create view frag_ops as select
if(type in ('Ordered index', 'Unique hash index'),
substring(fq_name from 1+locate('/', fq_name, 9)),
if(fq_name like '%BLOB%', left(fq_name, locate('_', fq_name)), fq_name))
as fq_name,
parent_fq_name,
type,
count(fragment_num) as frag_count,
sum(tot_key_reads) as tot_key_reads,
sum(tot_key_inserts) as tot_key_inserts,
sum(tot_key_updates) as tot_key_updates,
sum(tot_key_writes) as tot_key_writes,
sum(tot_key_deletes) as tot_key_deletes,
sum(tot_key_refs) as tot_key_refs,
round(log2(1+sum(tot_key_attrinfo_bytes))) as log_tot_key_attrinfo_bytes,
sum(tot_key_keyinfo_bytes) as tot_key_keyinfo_bytes,
sum(tot_key_prog_bytes) as tot_key_prog_bytes,
sum(tot_key_inst_exec) as tot_key_inst_exec,
sum(tot_key_bytes_returned) as tot_key_bytes_returned,
sum(tot_frag_scans) as tot_frag_scans,
sum(tot_scan_rows_examined) as tot_scan_rows_examined,
sum(tot_scan_rows_returned) as tot_scan_rows_returned,
sum(tot_scan_bytes_returned) as tot_scan_bytes_returned,
sum(tot_scan_prog_bytes) as tot_scan_prog_bytes,
sum(tot_scan_bound_bytes) as tot_scan_bound_bytes,
sum(tot_scan_inst_exec) as tot_scan_inst_exec,
# Ignore concurrent frag_scans, as all tests below expect them to be 0
# but this is not stable as LCP scans are counted as concurrent
# frag scans while they are running (but not counted as tot_frag_scans
# when they complete).
# Arguably it is useful to see a live LCP scan, and this test has no
# use for conc_frag_scans!=0, so avoid looking at it for determinism
#sum(conc_frag_scans) as conc_frag_scans,
sum(conc_qd_frag_scans) as conc_qd_frag_scans,
sum(tot_commits) as tot_commits
from ndbinfo.operations_per_fragment opf
where fq_name like 'ndbinfo%' or parent_fq_name like 'ndbinfo%'
group by table_id, type, opf.fq_name, parent_fq_name
order by ifnull(parent_fq_name, opf.fq_name), opf.fq_name;
# Create and populate test tables.
create table t1 (
a int not null,
b int not null,
c int not null,
d int not null,
t text not null,
primary key (b,a)
) engine=ndbcluster
partition by key(b,a) partitions 8;
create unique index ix1 on t1(d) using hash;
# A helper table for populating t1.
create table numbers(
i int
);
insert into numbers values (0),(1),(2),(3),(4),(5),(6),(7),(8),(9);
insert into t1 select x1.i+x2.i*10, 1, 1+x1.i+x2.i*10, 1+x1.i+x2.i*10,
repeat('x', 512)
from numbers as x1, numbers as x2;
drop table numbers;
create table t2 (
a int,
b int,
primary key(a) using hash)
engine = ndb
partition by key(a) partitions 8;
insert into t2 values (1, 2);
insert into t2 values (2, 3);
insert into t2 values (3, 1);
# Check counters before the test queries.
--query_vertical select * from frag_ops;
# Do a table scan with a pushed condition.
--replace_column 4 #
explain select count(*) from t1 where d<10;
select count(*) from t1 where d<10;
# Do an index scan.
--replace_column 4 #
explain select count(*) from t1 where b=1 and a<10;
select count(*) from t1 where b=1 and a<10;
# Do a PK read.
--replace_column 4 #
explain select * from t1 where b=1 and a = -5;
select * from t1 where b=1 and a = -5;
# Try to insert a duplicate key.
--error ER_DUP_ENTRY
insert into t1 values (1,1,1,1,'');
# Use replace to generate PK write operations.
replace t1 values (-1,-1,-1,-1,'');
replace t1 values (1,1,5,500,'');
# Do update and delete.
update t1 set d = -d, t = repeat('a', 300) where b=1 and a > 90;
delete from t1 where a%30=0;
# check_ndb_debug sets have_ndb_debug to 1 if the
# Ndb binaries have debug support built in
# This is needed for error-inserts etc.
source include/check_ndb_debug.inc;
# This error insert causes the next index scan from SPJ to be queued on node 2.
# This command may fail if error inserts is not available on this build,
# therefore we permit return code 255.
--error 0,255
--exec $NDB_MGM --no-defaults --ndb-connectstring="$NDB_CONNECTSTRING" -e "2 error 5084" >> $NDB_TOOLS_OUTPUT
# Do a pushed query that starts multiple index scans on each t1 fragment.
# That way the first scan (on node 2) will be queued until the next finishes.
--replace_column 4 #
explain select count(*) from t1 as x1 join t1 as x2 on x1.c=x2.b;
select count(*) from t1 as x1 join t1 as x2 on x1.c=x2.b;
# Check number of queued scans. Test for debug build
# to handle builds without error inserts (i.e. non-debug builds).
--disable_query_log
eval select
if ($have_ndb_debug,
sum(tot_qd_frag_scans),
4) as tot_qd_frag_scans
from ndbinfo.operations_per_fragment
where fq_name like 'ndbinfo%' or parent_fq_name like 'ndbinfo%';
--enable_query_log
# Check how these operations changed the counters.
--query_vertical select * from frag_ops;
# This query should give PK read operations with an interpreted program.
# (Multi-range read mapped to PK reads.)
--replace_column 4 #
explain select count(*)
from t2 as tx, t2 as ty
where tx.a in (1,3,5)
and ty.a = tx.b;
# The 'explain' above triggers an implicit 'analyze table'. That increments
# operations_per_fragment counters in a way that is endian-dependent.
# We therefore save the current values of the counters to isolate the
# effect of the query below.
create table save_counters as select * from frag_ops;
select count(*)
from t2 as tx, t2 as ty
where tx.a in (1,3,5)
and ty.a = tx.b;
create view changes as select
new.fq_name,
new.parent_fq_name,
new.type,
new.frag_count,
new.tot_key_reads - old.tot_key_reads,
new.tot_key_inserts - old.tot_key_inserts,
new.tot_key_updates - old.tot_key_updates,
new.tot_key_writes - old.tot_key_writes,
new.tot_key_deletes - old.tot_key_deletes,
new.tot_key_refs - old.tot_key_refs,
new.log_tot_key_attrinfo_bytes - old.log_tot_key_attrinfo_bytes,
new.tot_key_keyinfo_bytes - old.tot_key_keyinfo_bytes,
new.tot_key_prog_bytes - old.tot_key_prog_bytes,
new.tot_key_inst_exec - old.tot_key_inst_exec,
new.tot_key_bytes_returned - old.tot_key_bytes_returned,
new.tot_frag_scans - old.tot_frag_scans,
new.tot_scan_rows_examined - old.tot_scan_rows_examined,
new.tot_scan_rows_returned - old.tot_scan_rows_returned,
new.tot_scan_bytes_returned - old.tot_scan_bytes_returned,
new.tot_scan_prog_bytes - old.tot_scan_prog_bytes,
new.tot_scan_bound_bytes - old.tot_scan_bound_bytes,
new.tot_scan_inst_exec - old.tot_scan_inst_exec,
new.tot_commits - old.tot_commits
from frag_ops as new, save_counters as old
where new.fq_name=old.fq_name and new.fq_name like '%t2';
--query_vertical select * from changes
# Drop test views and tables.
drop view changes;
drop table save_counters;
drop view frag_ops;
drop table t1;
drop table t2;
#
# WL#7575 Remove ndbinfo's usage of other engine
# - the five lookup tables which resided in another engine
# was converted to virtual(aka. hardcoded) tables inside ha_ndbinfo
# - three new views was added to preserve backwards compatibility
# for those tabes which was changed to prefix ndb$
## Check that the new tables return data and exists.
select count(*) > 10 from ndb$blocks;
select param_number, param_name, param_description, param_default,
param_min, param_max from ndb$config_params
where param_name = "DataMemory";
select * from ndb$dict_obj_types where type_id < 3;
--disable_result_log
select * from ndb$dbtc_apiconnect_state;
select * from ndb$dblqh_tcconnect_state;
--enable_result_log
## Check that the new views return data and exists.
select block_number from blocks where block_name = "DBTC";
select param_name from config_params where param_number = 101;
select type_name from ndb$dict_obj_types where type_id = 3;
## All tables in ndbinfo/ should now be using engine=NDBINFO
select count(*) from information_schema.tables
where table_schema = 'ndbinfo' and engine != 'ndbinfo';
#
# Bug#16731538 MYSQLD CRITICAL FAILURE DURING ORDERED SELECT FROM NDBINFO.CLUSTER_OPERATIONS
# - ha_ndbinfo::estimate_rows_upper_bound() returned 0 rows
#
--disable_result_log
select * from ndbinfo.config_params order by param_number;
--enable_result_log
#
# Bug#11762750 TABLE NDBINFO.CONFIG_PARAMS SHOULD BE READ-ONLY (FOR NOW)
#
#
# Checking the table
--error ER_OPEN_AS_READONLY
UPDATE ndb$config_params SET param_number=1 WHERE param_name = "NoOfReplicas";
# Checking the view
--error ER_OPEN_AS_READONLY
UPDATE config_params SET param_number=1 WHERE param_name = "NoOfReplicas";
#
# WL#8703 ndbinfo config parameter tables
#
select node_id, param_name, config_param, config_value
from ndbinfo.config_values cv, ndbinfo.config_params cp
where cv.config_param = cp.param_number and
cp.param_name = 'MaxNoOfConcurrentTransactions'
order by node_id;
select node_id, param_name, config_value
from ndbinfo.config_values cv, ndbinfo.config_params cp
where cv.config_param = cp.param_number and
cp.param_name = 'NodeId'
order by node_id;
select node_id, concat(config_value / 1024 / 1024, 'M') as DataMemory
from ndbinfo.config_values
where config_param = 112 /* DataMemory */
order by node_id;
#
# wl#9819 ndbinfo.processes
#
# Because port number and host address vary in the test environment,
# we can not test the content of the service_URI column.
# (A) Normative tests of table and view definition
# (turn show_hidden on since otherwise columns
# of the table is supposed to be hidden)
set @@ndbinfo_show_hidden=TRUE;
desc ndb$processes;
set @@ndbinfo_show_hidden=default;
show create table processes;
# (B) Tests of table content.
# These results will change if the cluster config used when running
# this test case changes.
# Query the base table
SELECT reporting_node_id, node_id, node_type,
replace(replace(process_name,"-debug",""),".exe","") AS process_name
FROM ndb$processes
ORDER BY reporting_node_id, node_id;
# Query the view
#
# Note, on Windows mysqld processes will have an angel_process_id while
# on other platforms there will be none.
#
# The Windows angel_process_id will be reported in monitor_proc_id
# instead.
#
# On Windows the angel_proc_id will always be set to "null" for API.
# On other platforms monitor_proc_id will instead always be set to
# "not_null" for API.
#
# Test for node_type = 'API' is used since checking for process_name =
# 'mysqld' would be much more cumbersome, and the only API node running
# in this case are the mysqld servers.
set @windows = convert(@@version_compile_os using latin1)
in ('Win32', 'Win64', 'Windows');
SELECT node_id,
node_type,
replace(replace(process_name,"-debug",""),".exe","") AS process_name,
if(process_id is null, "null", "not_null") as proc_id,
if(@windows AND node_type = 'API',
"null",
if(angel_process_id is null, "null", "not_null")) as angel_proc_id,
if(NOT @windows AND node_type = 'API',
"not_null",
if(angel_process_id is null, "null", "not_null")) as monitor_proc_id
FROM processes
ORDER BY node_id;
#
# wl#10147 ndbinfo.config_nodes
#
# (A) Normative tests of table and view definition
# (turn show_hidden on since otherwise columns
# of the table is supposed to be hidden)
set @@ndbinfo_show_hidden=TRUE;
desc ndb$config_nodes;
set @@ndbinfo_show_hidden=default;
select view_definition, is_updatable, security_type
FROM information_schema.views
WHERE table_schema = 'ndbinfo' AND table_name = 'config_nodes';
# (B) Tests of table content.
# These results will change if the cluster config used when running
# this test case changes.
SELECT * from ndb$config_nodes order by reporting_node_id, node_id;
# (C) Tests of view: SELECT DISTINCT, removing reporting_node_id
SELECT node_type, count(*),
GROUP_CONCAT(node_id order by 1 asc separator " ") ids
FROM config_nodes group by node_type order by node_type;
--echo #
--echo # Bug#26048272 IMPLEMENT NDBINFO TABLE FOR LOOKUP OF ERROR CODES AND MESSAGES
--echo #
# Check that row is returned for NDB error 920
select *
from ndbinfo.error_messages
where error_code = 920;
# Check that row is returned for NDB exit code 2304
select *
from ndbinfo.error_messages
where error_code = 2304;
# Check that row is returned for MgmApi error code 5001
select *
from ndbinfo.error_messages
where error_code = 5001;
# Check that there are at last 700 rows in the error_message table, don't
# use a hard limit as it will grow when someone add a new error message
select count(*) > 700
from ndbinfo.error_messages;
--remove_file $NDB_TOOLS_OUTPUT