[MCOL-5136] error on projection for oid 1001 lbid 0: input rids 80, output rids 79#012: Restarted a syscat job 120 times, bailing Created: 2022-06-14  Updated: 2023-11-17

Status: Open
Project: MariaDB ColumnStore
Component/s: None
Affects Version/s: 5.6.5, 6.2.2
Fix Version/s: Icebox

Type: Task Priority: Major
Reporter: Edward Stoever Assignee: Unassigned
Resolution: Unresolved Votes: 0
Labels: None


 Description   

To support CS0401767 I have attempted to reproduce the error:

May 22 06:55:43 demartdb01 messagequeue[76949]: 43.348484 |0|0|0| W 31 CAL0000: MessageQueueClient::write: error writing 5202 bytes to IOSocket: sd: 217 inet: 127.0.0.1 port: 8601. Socket error was InetStreamSocket::write error: Broken pipe -- write from InetStreamSocket: sd: 217 inet: 127.0.0.1 port: 8601
 
May 22 06:57:43 demartdb01 PrimProc[77317]: 43.279036 |0|0|0| C 28 CAL0000: /home/jenkins/workspace/ES/RPM-Builds/arch/x86_64/env/build/os/rhel-7/padding_for_CPACK_RPM_BUILD_SOURCE_DIRS_PREFIX_ON_ES_BACKUP_DEBUGSOURCE/storage/columnstore/columnstore/primitives/primproc/rtscommand.cpp (token column) error on projection for oid 1001 lbid 0: input rids 80, output rids 79#012: Restarted a syscat job 120 times, bailing

From what I understand, error occurs when rtscommand.cpp calls bppseeder.cpp. Source code is identifiable from the keyword "bailing".

Some of the steps I have tried to recreate this error:

  • intentionally corrupt Columnstore files such as any or many of the "FILE000.cdf"
  • intentionally corrupt files, one by one, in /var/lib/columnstore/data1/systemFiles/dbrm
  • change high watermark for oid 1001 to very high numbers until it breaks

None of these scenarios worked to reproduce the error. To understand the error and help prevent it, we would like to reproduce it and document a way to fix it, if possible.
Generated at Thu Feb 08 02:55:36 UTC 2024 using Jira 8.20.16#820016-sha1:9d11dbea5f4be3d4cc21f03a88dd11d8c8687422.