MySQL Bugs: #55045: Queries stop for excessive time when master data node stopped gracefully

Bug #55045	Queries stop for excessive time when master data node stopped gracefully
Submitted:	7 Jul 2010 8:04	Modified:	28 Jul 2016 17:36
Reporter:	Andrew Morgan	Email Updates:
Status:	Closed	Impact on me:	None
Category:	MySQL Cluster: Cluster (NDB) storage engine	Severity:	S3 (Non-critical)
Version:	mysql-5.1-telco-7.1	OS:	Any
Assigned to:	MySQL Verification Team	CPU Architecture:	Any
Tags:	7.1

Description:
When gracefully stopping a ndbd process, queries are stopped for a number of seconds. This only seems to happen when the data node is tagged as being the master for the node group.

Most common error is 4030 but also seen 4006 & 4029. Mostly done on a VM running Ubuntu but confirmed on native Fedora.

Problem initially experienced with Cluster created using MySQL Cluster Manager but seems to be the same for Clusters created by hand.

How to repeat:
Setup = tiny database, 1 single column table with 6 rows. 2 mysql clients connected to the same mysqld.

Stop the data node that is shown as the master.

Keep repeating a SELECT ALL in mysql client 1 until it hangs.

Wait a couple of seconds.

Try SELECT ALL from mysql client 2 and confirm that it hangs for a second or so.

Config files (created by MCM)

Attachment: no_mcm.tar.gz (application/x-gzip, text), 323 bytes.

Config files (created using MCM)

Attachment: mcm.tar.gz (application/x-gzip, text), 506 bytes.

Hi,

I can't repeat this...

I'm using
- debug compiled
- create table t1 (a int) engine = ndb;
- insert into t1 values (1), (2), (3), (4), (5), (6);
- while true; do date; mysql -uroot test -e "select * from t1"; done

no hangs what so ever...

I saw the same results as Jonas when using "mysql -e" for each query.

The problem comes when using the mysql CLI in interactive mode (perhaps because it doesn't create a new connection each time?). Note that Mario was experiencing this when using JDBC. I get the following after the command pauses for a few seconds...

mysql> select * from t1;
ERROR 1297 (HY000): Got temporary error 4030 'Transaction aborted due to node shutdown' from NDBCLUSTER
mysql> 100831 12:55:10 [Note] NDB Binlog: Node: 3, down, Subscriber bitmask 00

7.4.11 - this is definitively solved 
 - tried interactive
 - tried non interactive
 - tried from odbc directly
 - tried from odbc pool
 - tried from capi

no "freeze", 2 node cluster, stopping master node, starting it back, stopping new master node, starting it back, stopping slave node, starting it back up .. no "freeze" no delays at all on the mysql side

identical behavior with 4 node cluster

marking closed