On Thu, Mar 28, 2019 at 12:07 PM Simon Riggs <simon@2ndquadrant.com> wrote:
> Seems more like our own labelling of the phases is responsible for that, rather than it being a specific problem. The
numberingshould reflect the ordinal executed step number. So if a VACUUM has required two sets of index scanning, the
heapscan phase (normally phase 3) should be labelled phase 6 when it occurs the second time, rather than "phase 3
again,doh" which clearly doesn't work.
That would not be too simple to do with the infrastructure we have
available, I think. Also, Alvaro's showed phase names like '3 of 8',
but if you regarded each set of index scans as a separate phase rather
than a repetition of a phase that had already happened, you wouldn't
know whether there were going to be 8 phases in total or some other
number, because you don't know how many times you're going to scan the
indexes.
I suggest that it makes sense to leave the phase numbers out of this
commit. If someone wants to make a proposal for adding phase numbers
to the various commands that now support progress reporting as a
separate commit, then that can be discussed separately.
BTW, if we are going to do that, it might be best to put then in
separate view columns rather than making them part of the phase names.
Somebody might want to look for those phase names using SQL or
client-side logic and not have the logic get broken if we renumber the
phases.
--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company