[MCOL-5218] Fix mcs-loadbrm service start while reinstall. Created: 2022-09-13  Updated: 2022-10-19  Resolved: 2022-09-26

Status: Closed
Project: MariaDB ColumnStore
Component/s: cmapi, installation
Affects Version/s: 22.08.1
Fix Version/s: 22.08.2

Type: Task Priority: Blocker
Reporter: Daniel Lee (Inactive) Assignee: Alan Mologorsky
Resolution: Fixed Votes: 0
Labels: None

Assigned for Review: Leonid Fedorov Leonid Fedorov

 Description   

For rpm upgrades, if the user uninstall old MariaDB packages and reinstall new 22.08.1 packages, the system catalog would become empty. -This is because cmapi packages gets installed last, and the ColumnStore engine packages requires CMAPI. Should we make CMAPI as a requirement for the engine?

The reason is in mcs-loadbrm.py script.
Steps to fix:

  • remove external requests dependency
  • make a fallback for primary node detection that should work without cmapi installed

 Comments   
Comment by Leonid Fedorov [ 2022-09-16 ]

dleeyh please see the builds of PR made by Allan
https://cspkg.s3.amazonaws.com/index.html?prefix=develop/pull_request/5593/10.9/amd64/rockylinux8/

and
https://cspkg.s3.amazonaws.com/index.html?prefix=develop/pull_request/5593/10.9/amd64/centos7/

Could you please test them against upgrade issue&

Comment by Daniel Lee (Inactive) [ 2022-09-16 ]

Please note that the links above are for server 10.9. We should be using the 10.6-enterprise instead. I did test both and both failed with the following errors. defaultTest.sql is a sanity SQL test script which I run on every installation/upgrade test.

Tested on rocky 8 using RPM package installation and upgrade (rpm -e and rpm -ivh)

After upgrade is completed

    s2pm1: executing defaultTest.sql
 
    s2pm1: Fri Sep 16 16:59:46 UTC 2022
 
    s2pm1: ERROR 1815 (HY000) at line 9: Internal error: CAL0001: Insert Failed:  Lost connection to Write Engine Server while updating SYSTABLES
 
    s2pm1: ERROR 1815 (HY000) at line 10: Internal error: CAL0001: Insert Failed:  Lost connection to WriteEngineServer on pm2
 
    s2pm1: ERROR 1815 (HY000) at line 11: Internal error: Lost connection to ExeMgr. Please contact your administrator
 
    s2pm1: ERROR 1815 (HY000) at line 12: Internal error: InetStreamSocket::connect: connect() error: Connection refused to: InetStreamSocket: sd: 67 inet: 10.0.0.35 port: 8601
 
    s2pm1: ERROR 1815 (HY000) at line 13: Internal error: InetStreamSocket::connect: connect() error: Connection refused to: InetStreamSocket: sd: 67 inet: 10.0.0.35 port: 8601

When selecting from the system catalog systable table

Type 'help;' or '\h' for help. Type '\c' to clear the current input statement.
 
 
 
MariaDB [calpontsys]> select * from systable;
 
ERROR 1815 (HY000): Internal error: InetStreamSocket::readToMagic(): I/O error1: rc-1; poll signal interrupt ( POLLHUP POLLERR )
 
MariaDB [calpontsys]> exit

Installation and upgrade using repo worked fine, as reported previously.

Comment by alexey vorovich (Inactive) [ 2022-09-30 ]

fixed

Generated at Thu Feb 08 02:56:14 UTC 2024 using Jira 8.20.16#820016-sha1:9d11dbea5f4be3d4cc21f03a88dd11d8c8687422.