Robert Haas wrote:
> On Thu, Jul 16, 2015 at 9:41 AM, Fujii Masao <masao.fujii@gmail.com> wrote:
> > Here are some minor comments:
> >
> > + ereport(LOG,
> > + (errmsg("ignoring \"%s\" file because no
> > \"%s\" file exists",
> > + TABLESPACE_MAP, BACKUP_LABEL_FILE),
> > + errdetail("could not rename file \"%s\" to
> > \"%s\": %m",
> > + TABLESPACE_MAP, TABLESPACE_MAP_OLD)));
> >
> > WARNING is better than LOG here because it indicates a problematic case?
>
> No, that's not the right distinction. Remember that, when sending
> messages to the client, WARNING > LOG, and when sending messages to
> the log, LOG > WARNING. So messages that a user is more likely to
> care about than the administrator should be logged at WARNNG; those
> that the administrator is more likely to care about should be LOG. I
> think LOG is clearly the appropriate thing here.
Right. Now that we have errors marked with criticality, it will be
pretty obvious what message is indicating a system problem rather than
just a problem that can be ignored without taking any action.
... oh, actually we don't have the criticality marking yet, do we.
I think we need to fix that at some point. (Some guys have said to grep
the log for certain SQLSTATE codes, but, uh, really?)
--
Álvaro Herrera http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services
