Comment 27 for bug 1931737

All,

I tested this branch on a copy of a production database. On Postgres 10. The database was pg_dumped from pg9.6, then restored onto a pg10 instance. Then upgraded EG3.7 -> EG3.9 with this patch. I performed the sideloader and then pingest. Before I started the ingest, I noted the size of the database on disk: 16GB. The first time I used pingest.pl, I did not pass the --delay-symspell switch. The DB grew 2GB, to 18GB. Funny enough, there were no deadlocks without the switch. Then I ran it again with the switch. When it finished, it grew to 20GB. In conclusion, it seems that the database grew by 25% of it's original size on disk. (4 / 16) * 100 = 25%. That's not great, however, this database had previously not had symspell enabled. This test database does not have an external script to deal with the WAL logs, so they are getting rolled off immediately. Not soaking up the disk. The pg_wal folder is 529MB.

FWIW, I've given my sign-off, I did find a missing semicolon which I committed here as well:

https://git.evergreen-ils.org/?p=working/Evergreen.git;a=shortlog;h=refs/heads/user/blake/lp1931737-dym-deadlock-plus-pingest-rebase-signoff