| From: | Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> |
|---|---|
| To: | Justin Pryzby <pryzby(at)telsasoft(dot)com> |
| Cc: | Ilya Gladyshev <ilya(dot)v(dot)gladyshev(at)gmail(dot)com>, Matthias van de Meent <boekewurm+postgres(at)gmail(dot)com>, Alvaro Herrera <alvherre(at)alvh(dot)no-ip(dot)org>, Tomas Vondra <tomas(dot)vondra(at)enterprisedb(dot)com>, pgsql-hackers(at)lists(dot)postgresql(dot)org |
| Subject: | Re: Progress report of CREATE INDEX for nested partitioned tables |
| Date: | 2023-03-25 19:43:32 |
| Message-ID: | 2773247.1679773412@sss.pgh.pa.us |
| Views: | Raw Message | Whole Thread | Download mbox | Resend email |
| Thread: | |
| Lists: | pgsql-hackers |
I pushed 0001 with some cosmetic changes (for instance, trying to
make the style of the doc entries for partitions_total/partitions_done
match the rest of their table).
I'm not touching 0002 or 0003, because I think they're fundamentally
a bad idea. Progress reporting is inherently inexact, because it's
so hard to predict the amount of work to be done in advance -- have
you ever seen a system anywhere whose progress bars reliably advance
at a uniform rate? I think adding assertions that the estimates are
error-free is just going to cause headaches. As an example, I added
a comment pointing out that the current fix won't crash and burn if
the caller failed to lock all the child tables in advance: the
find_all_inheritors call should be safe anyway, so the worst consequence
would be an imprecise partitions_total estimate. But that argument
falls down if we're going to add assertions that partitions_total
isn't in error.
regards, tom lane
| From | Date | Subject | |
|---|---|---|---|
| Next Message | Tom Lane | 2023-03-25 19:58:51 | Re: Infinite Interval |
| Previous Message | Jeff Davis | 2023-03-25 19:16:35 | Re: Non-superuser subscription owners |