[MCOL-3725] Stopping ColumnStore via mcsadmin shutdownsystem during bulk injection via mcsapi makes the table unuseable Created: 2020-01-15  Updated: 2021-02-21  Resolved: 2021-02-21

Status: Closed
Project: MariaDB ColumnStore
Component/s: N/A
Affects Version/s: 1.4.2
Fix Version/s: N/A

Type: Bug Priority: Critical
Reporter: Jens Röwekamp (Inactive) Assignee: Unassigned
Resolution: Fixed Votes: 1
Labels: SkySQLMVP
Environment:

MariaDB ColumnStore single node on CentOS 7 with S3 storage bucket on GCP (Virtual machine not pod)


 Description   

Forcefully stopping MariaDB ColumnStore via mcsadmin shutdownsystem while a bulk injection via mcsapi or related tools is executed makes the target table of the bulk injection unuseable.

No more data can be injected after the system is restarted via mcsadmin startystem.

MariaDB [test]> insert into mcsapi_load_test values (23, 42, 2342);
 
ERROR 1815 (HY000): Internal error: CAL0001: Insert Failed:  Error getting extent information for table mcsapi_load_test



 Comments   
Comment by Zdravelina Sokolovska (Inactive) [ 2020-01-15 ]

jens.rowekamp when you are issuing 'mcsadmin shutdownsystem '
that should not happen since those are table locks.
if force mcsadmin shutdownsystem while those are table locks
rollback should happen first , if I am not wrong that was fixed in columnstore time ago.

Anyway, the implemented shutdown hooks will restart the Columnstore Pod after shutdown mcs.
so that's the same with #DBAAS-2008 ;
from Kubernated point of view Pod restarting is the use case thought rather than force mcs down

Comment by Todd Stoffel (Inactive) [ 2021-02-21 ]

Obsolete

Generated at Thu Feb 08 02:44:58 UTC 2024 using Jira 8.20.16#820016-sha1:9d11dbea5f4be3d4cc21f03a88dd11d8c8687422.