Allow migrating blob/clob when it is known that the column size is relatively small

[amit-yb]

No description provided.

[amit-yb]

@rahulb-yb Please update this ticket with whatever decision we took about blob/clob support.

[rahulb-yb]

For the documentation:
BLOBs and CLOBs are not native data types in PG- they are closest represented by the native bytea and text datatypes respectively.
While BLOBs and CLOBs have better support for storage of data, bytea and text only allow upto 1 GB of data per cell, meaning that if the data rows we attempt to import contain >1GB of data, it would fail

As an alternative, PG provides support for LOB which are capable of holding BLOB/CLOB datatypes. This, however, is not yet supported in YB, making this alternative infeasible.

Finally, BLOB and CLOB data types can contain a lot of more data per cell as compared to most other native data types, which can cause timeout errors when combined with our logic for parallelism of import data- the size can potentially be greater by an order of upto 2^25 compared to most other data types, making it infeasible to attempt to migrate it alongside the rest of the data. Our current build already ignores importing the BLOB and CLOB data, but keeps the schema preserved, allowing the user to import their data onto the preserved schema by any means they wish

[shubham-yb]

From a recent experiment it was observed that CLOB datatype row values are migrated but the upper limit of the datafile was seen to be 243 MB.
Any greater than that and the migration errors out.
Further details are in this sheet