[MCOL-688] Delete failed, several errors Created: 2017-04-28  Updated: 2017-08-09  Resolved: 2017-08-09

Status: Closed
Project: MariaDB ColumnStore
Component/s: ExeMgr
Affects Version/s: 1.0.8
Fix Version/s: Icebox

Type: Bug Priority: Major
Reporter: Seungmin Yang Assignee: Unassigned
Resolution: Duplicate Votes: 0
Labels: community
Environment:

RHEL 6.6



 Description   

I installed MariaDB ColumnStore 1.0.8. I use 1 PM and the storage is internal. I'm dealing with bulk data. For instance, 'test_log' table has 100,000,000 rows.

It succeeded to delete 100,000 rows.

MariaDB [input_test]> delete from test_log where stime < '2017-03-31 00:00:00' limit 100000;
Query OK, 100000 rows affected (9.67 sec)

However, when I tried deleting 1,000,000 rows from test_log table, it shows an error as follows:

MariaDB [input_test]> delete from test_log where stime < '2017-03-31 00:00:00' limit 1000000;
ERROR 1815 (HY000): Internal error: ClientRotator caught exception: InetStreamSocket::write error: Broken pipe – write from InetStreamSocket: sd: 19 inet: 127.0.0.1 port: 8601

I performed the same query again, it succeeded!

MariaDB [input_test]> delete from test_log where stime < '2017-03-31 00:00:00' limit 1000000;
Query OK, 1000000 rows affected (1 min 28.25 sec)

Once I deleted 10,000,000 rows, another error occurred.

MariaDB [input_test]> delete from test_log where stime < '2017-03-31 00:00:00' limit 10000000;
ERROR 2013 (HY000): Lost connection to MySQL server during query

After the error occurred, I couldn't perform other queries with the table because it's locked.

MariaDB [input_test]> delete from test_log where stime < '2017-03-31 00:00:00' limit 10000000;
ERROR 1815 (HY000): Internal error: IDB-2009: Unable to perform the delete operation because DMLProc with PID 21714 is currently holding the table lock for session 2.

MariaDB [input_test]> truncate table test_log;
ERROR 1815 (HY000): Internal error: CAL0009: Truncate table failed: IDB-2009: Unable to perform the DMLProc operation because 21714 with PID 2 is currently holding the table lock for session .

Could you let me know what the problem is? Why do these errors occur? And how can I fix it?

Thank you.



 Comments   
Comment by Andrew Hutchings (Inactive) [ 2017-04-28 ]

The problem is likely MCOL-529. Try the following as root:

/sbin/sysctl net.ipv4.ip_local_port_range="1024 65000"
/sbin/sysctl net.ipv4.tcp_fin_timeout=30
/sbin/sysctl net.ipv4.tcp_tw_recycle=1
/sbin/sysctl net.ipv4.tcp_tw_reuse=1

Then use "mcsadmin restart" to restart the ColumnStore server.

Please let us know whether this does or does not resolve the problem for you.

Note: these will reset when you reboot, you can make them permanent in /etc/sysctl.conf

Comment by David Thompson (Inactive) [ 2017-06-05 ]

Hi, can you confirm if the workaround or using version 1.0.9 resolved this issue for you?

Comment by David Thompson (Inactive) [ 2017-08-09 ]

Please re-open if this is still a problem.

Generated at Thu Feb 08 02:23:04 UTC 2024 using Jira 8.20.16#820016-sha1:9d11dbea5f4be3d4cc21f03a88dd11d8c8687422.