since bsos INTERLEAVE's w/ DELETE CASCADE
- persist unique_key_filter across writes
- fix new bundling of of bso_values for inserting bsos
- add TODO for fixing user_collections' modified time
use `ms_delay` to pause between spanner transaction `--readchunk`s. This allows
some primative throttling for feeding spanner data.
Reminder: `readchunk` sets the max number of items to try to write per chunk
to spanner in any given transaction, default value 1000.
The `--user_percent` option will divvy up the users into blocks and
move the specified block. It takes an option formatted as
"block#:percentage". Block numbers are 1 based. For example,
--user_percent=2:33 will divide the total distinct users into
non-overlapping blocks of approximately 33%, and then move the second
block (e.g. the 33-65th users in the list). Extra users that may not be
evenly divided into percentage blocks will be appended to the last
block. (e.g. for `--user_percent=3:33`, users 66-99 would be copied
over, a total of 34 users)
Issue #407
* feat: more user_migration stuff
* create script to move users by node directly
* moved old scripts to `old` directory (for historic reasons, as well as
possible future use)
* cleaned up README
* try to solve the `parent row` error
an intermittent error may be responsible from one of two things:
1) a transaction failure resulted in a premature add of the unique key
to the UC filter.
2) an internal spanner update error resulting from trying to write the
bso before the user_collection row was written.
* Added "fix_collections.sql" script to update collections table to add
well known collections for future rectification.
* returned collection name lookup
* add "--user" arg to set bso and user id
* add `--dryrun` mode
