On May 24, 2020, at 7:24 AM, Stephen J. Turnbull <turnbull.stephen.fw@u.tsukuba.ac.jp> wrote:
Mark Sapiro wrote:
On 5/19/20 7:25 PM, Mark Dadgar wrote:
For reference my pgloader script looks like:
load database from sqlite:///var/lib/dbconfig-common/sqlite3/mailman3-web/mailman3web.db into postgresql://mailman3:XXREDACTEDXX@localhost/mailman
with include drop, create tables, create indexes, reset sequences, no truncate
Have you tried this without indexes and then run the migrate script? It may re-create them as part of the schema upgrade check.
That was my immediate guess. I'm not an SQL expert, but I believe that sequences and indexes are different entities, they're both optimizations that you don't need (ie, ignoring performance), and implementations differ on how they're implemented. Eg, if sqlite implements them as tables, while PostgreSQL has a special data structure, you could run into this kind of behavior. I'm pretty sure that these will be recreated appropriate to the new database when the schema validator is done.
But don't trust me, and *do* keep a backup. ;-)
😁
I did try it without indexes and it started showing the same warnings.
I will give this another try and let it run to completion. I was hesitant to do that because the core database is 7.5 GB and the web database is 35 GB so there are only so many copies I can keep hanging around on the cloud volume.
Since the data is mostly text, I am hoping that postgres will compress the bejeesus out of it.
- Mark
mark@pdc-racing.net | 408-348-2878