[Building Sakai] Problem with clustering??

[Seth Theriault]

On Wed, Sep 26, 2012 at 6:33 PM, Matthew Jones <matthew at longsight.com> wrote:

> The intention was to fix cluster problems when daylight savings time
> changed, but it seemed like it would break more than it would fix. He
> assured me that there was research done that all of this was safe, but I
> don't have any way to test on Oracle (and not much vested interest in
> Oracle) so accepted this.

Looking back at KNL-725, I am the source of the research and therefore
probably the problems that have occurred. I clearly missed the
subtleties involved and opnly verified whether the conversion from
DATE to a TIMESTAMP variant would be OK for the data preservation.

This was a critical error on my part.

> It's very likely it should have been converted to TIMESTAMP WITH LOCAL TIME
> ZONE. Then it would have been at least equal to Mysql.

This is correct and goal of alignment is good. The fix is a different
story, especially if you have older data and you are expecting UTC.
I'd have to look more closely but it might be possible to convert the
affected data to UTC and then alter the column type (might need
external temporary tables for this). Then again, in a large
installation, this could affect millions of rows.

On a different note, I have scheduled Sakai restarts on the "DST
Sundays" for the last few years to workaround this problem in Oracle.

Seth