[Libreoffice-qa] How it went with one of the files used in last weeks bug hunting session
drewjensen.inbox at gmail.com
Sun Jun 3 01:43:58 UTC 2018
Howdy Tamás, list;
Last weekend during the 6.1 bug hunting session I prattled on, over on the
QA irc channel, about running the base migration assistant against an odb
file holding ~318 thousand records in an embedded hsql database.
Thought I'd post a recap here, with an added bit from today.
Before running the migration assistant;
Ran SHUTDOWN COMPACT against the original hsql odb which then had a 32.8
Meg disc image.
- pass one started
At the 4 hr and 30 minute mark I stopped the soffice process because I had
Did the following;
Dropped all indexes on all tables, recalling this schema is a reporting
dataset generated from a larger transaction set so lots of indexing on a
smaller number of tables.
With all indexes gone and cleaned up the disc image was down to 18.5 Megs.
- pass two
Can't say exactly how long it took, as it was late and after seeing this
change; during the first run the soffice process was writing data to disc
at a steady ~500K Bi/s,
during the second, without the indexs, the throuput went to ~1.2M Bi/s;
I went to bed.
In the morning and the migration complete the disc image was now 24.8 Megs
Picked that migrated file up again today adding back all, exact, indexing
and then jettisoned the old hsql files from the ODB.
Final disc image, 6.4 Megs.
BTW, the schema was advantageous in that it holds predominantly numeric
data, all character data in varchar or char fields, no blob data at all. It
included two text table definitions, which it dealt with properly by
ignoring them. And a view which, unfortunately, referenced one of those -
but the error again didn't stop the data load.
Could be happier with the results here.
My take away, how did I forget to check the indexing? (i'm getting
forgetful in my golden years..;)
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Libreoffice-qa