[MCOL-309] Support method to report on data set size Created: 2016-09-23  Updated: 2023-10-26  Resolved: 2016-11-29

Status: Closed
Project: MariaDB ColumnStore
Component/s: ?
Affects Version/s: None
Fix Version/s: 1.0.5

Type: New Feature Priority: Critical
Reporter: David Thompson (Inactive) Assignee: Daniel Lee (Inactive)
Resolution: Fixed Votes: 0
Labels: None

Issue Links:
Relates
relates to MCOL-406 Stored procedures required for I_S ta... Closed
Sprint: 2016-21, 2016-22, 2016-23

 Description   

There should be a means to determine storage actually used by columnstore by:

  • all
  • schema
  • schema and table

This should report by:

  • schema
  • tablename,
  • columnName
  • dataType
  • columnWidth
  • dbroot
  • partition
  • segment
  • filename
    -size (in GB compressed).
  • Uncompressed size would be desirable if possible too.

Information_Schema tables (Table and Column etc) likely would be the best and standard ways to support this, but not all of this information may be representable in that form. What can be populated in infoschema should be done and then we can review an approach for most easily capturing the remaining attributes as either a proc, script, or mcsadmin command.

 Comments   
Comment by David Thompson (Inactive) [ 2016-10-25 ]

Should look at info_schema as maybe the best approach but if not implement as an mcsadmin function e.g. runSizeReport. If the latter i think simple argument capability of (both optional):

  • Database name (including wildcard)
  • Table name (including wildcard)
    would suffice. The output should be something that is easily parsable and cut/pastable in case someone wants to run further shell analytics and / or import to a spreadsheet with no manual editing.
Comment by Andrew Hutchings (Inactive) [ 2016-11-07 ]

I first believed "calpontsys" might be the best place for this. Looking at historical information this database is used as a replacement for information_schema specifically for InfiniDB/ColumnStore tables. It is implemented in an InfiniDB/ColumnStore specific way and already contains some of what we need for this feature.

Digging deeper I find the calponsys is not fully implemented and the name doesn't fit the general direction of ColumnStore. I'll implement as information_schema and at some point we should retire calpontsys.

Comment by David Thompson (Inactive) [ 2016-11-07 ]

If we can make info schema work without too much contortions i think that's the best choice.

Comment by Andrew Hutchings (Inactive) [ 2016-11-14 ]

Requesting review from Hall, but dthompson might want to take a peek too.

Suggest it comes back to me before closing so that I can add documentation.

Comment by David Thompson (Inactive) [ 2016-11-29 ]

Closing this portion of the work for 1.0.5.

Generated at Thu Feb 08 02:20:05 UTC 2024 using Jira 8.20.16#820016-sha1:9d11dbea5f4be3d4cc21f03a88dd11d8c8687422.