Bug #35463 Cluster: BKA: CGE quiries hang mysqld demon
Submitted: 20 Mar 2008 18:27 Modified: 11 Feb 2009 14:37
Reporter: Jonathan Miller Email Updates:
Status: Closed Impact on me:
None 
Category:MySQL Cluster: Cluster (NDB) storage engine Severity:S2 (Serious)
Version:mysql-6.0-BKA OS:Linux (64)
Assigned to: Jan Wedvik CPU Architecture:Any

[20 Mar 2008 18:27] Jonathan Miller
Description:
In trying to create baselines and in trying to verify bug 33949 using the CGE benchmark, I keep hitting MySQLD hangs. In the queries listed below, the server will spike to 100% or better and never return. 

In SC Query (1) if I remove the where and order by conditions, the mysqld does not hang, but no data sets are ever returned. With the conditions removed, I can kill that application and the server returns to normal 

With the conditions applied, the MySQLD will spike to 104% and just stay there. Even after killing the test application, the server continues to stay in a spin.
14054 ndbdev    15   0 73508  12m 6268 S  104  0.1   3:03.26 mysqld

SC Querys (All)

When you run all 5 queries at one time, the mysqld demon will spike to 199% CPU and just hang there. Again in this case killing the test application has no affect on the MySQLD demon.
15772 ndbdev    15   0  356m  46m 6388 S  198  0.3  21:41.86 mysqld

NP Query (1) was passing a few days ago, but it to hangs the mysqld demon much like SC_Q1.
 

How to repeat:
Using CGE benchmark

 /space/cluster_rep_auto/tests/cge/mysqlchoke ndbXX:15001 sc /space/cluster_rep_auto/tests/cge/sql/sc/sc_query_1.sql 1 1 100000 0 500 1 1

 /space/cluster_rep_auto/tests/cge/mysqlchoke ndbXX:15001 sc /space/cluster_rep_auto/tests/cge/sql/sc/np_query_1.sql 1 1 100000 0 500 1 1

/space/cluster_rep_auto/tests/cge/mysqlchoke ndbXX:15001 sc /space/cluster_rep_auto/tests/cge/sql/sc/all_sc_queries.sql 20 1 100000 0 500 1 1

[atrt]
basedir=/data0/cr_autotest/run-reg-mysql-6.0-ndb-BKA/run
baseport=15000
clusters= .master

[ndb_mgmd]

[mysqld]
skip-grant-tables
skip-innodb
slave-allow-batching
ndb_use_exact_count=0

[cluster_config]
MaxNoOfSavedMessages = 1000

[cluster_config.master]
NoOfReplicas = 2
DataMemory = 4000M
IndexMemory = 400M
RedoBuffer=64M
NoOfFragmentLogFiles = 100
MaxNoOfConcurrentOperations = 250000
MaxNoOfLocalOperations = 275000
MaxNoOfConcurrentIndexOperations = 20000
MaxNoOfAttributes=2048
MaxNoOfOrderedIndexes=512
MaxNoOfUniqueHashIndexes=512
DiskPageBufferMemory=64MB
LockPagesInMainMemory=1
DiskCheckpointSpeed=16M

ndb_mgmd = ndb19
ndbd = ndbx9,ndbx0,ndbx9,ndbx0
mysqld = ndbx9
ndbapi=  ndbx8,ndbx8

[cluster_config.ndbd.1.master]
FileSystemPath=/data1/

[cluster_config.ndbd.2.master]
FileSystemPath=/data1/

[cluster_config.ndbd.3.master]
FileSystemPath=/data2/

[cluster_config.ndbd.4.master]
FileSystemPath=/data2/

#
# Generated by atrt
# Fri Mar 14 19:45:14 2008

[mysql_cluster.master]
ndb-connectstring= ndb1x:15000

[cluster_config.ndb_mgmd.1.master]
PortNumber= 15000

[mysqld.1.master]
datadir= /data0/cr_autotest/run-reg-mysql-6.0-ndb-BKA/run/mysqld.1
socket= /data0/cr_autotest/run-reg-mysql-6.0-ndb-BKA/run/mysqld.1/mysql.sock
port= 15001
ndb-connectstring= ndb19:15000
ndbcluster

[client.1.master]
host= ndbx9
socket= /data0/cr_autotest/run-reg-mysql-6.0-ndb-BKA/run/mysqld.1/mysql.sock
port= 15001
[18 Dec 2008 8:36] Jonas Oreland
jeb, 

lots of work has been done (by igor) on this
can you retest this bug on latest 6.0-ndb (and/or 60-bka-preview)

/Jonas
[11 Feb 2009 14:37] Jonathan Miller
Passing in the mysql-6.0-telco-6.4 clone