Обсуждение: WAL oddities (8.0.3)

First, are WAL files not allowed to end in FF?

I was looking at the logs and it jumps straight from
0000000100000194000000FE to 000000010000019500000000.

Yet other times it seems to end in an F: 0000000100000194000000EF.


Second, we have log archival enabled and the system it was archiving to
was unavailable. Once that system was brought back, everything resumed
appropriately and the files were copied over.

However, now I'm left with a pg_xlog directory that is about 7 GB in
size. All of the files but the most recent has a corresponding
archive_status/$FILE.done file. Will PostgreSQL eventually remove most
of these unnecessary files or am I stuck with them?

-- 

Rod Taylor <pg@rbt.ca> writes:
> First, are WAL files not allowed to end in FF?

I recall that there's some wastage of addressing space, but I don't
recall the exact reasoning.  You can probably find a comment about it
in xlog.c or nearby.

> However, now I'm left with a pg_xlog directory that is about 7 GB in
> size. All of the files but the most recent has a corresponding
> archive_status/$FILE.done file. Will PostgreSQL eventually remove most
> of these unnecessary files or am I stuck with them?

I'd have thought the next checkpoint would zap 'em.
        regards, tom lane

> > However, now I'm left with a pg_xlog directory that is about 7 GB in
> > size. All of the files but the most recent has a corresponding
> > archive_status/$FILE.done file. Will PostgreSQL eventually remove most
> > of these unnecessary files or am I stuck with them?
> 
> I'd have thought the next checkpoint would zap 'em.

I waited about an hour before sending this, but I guess a checkpoint
hadn't yet run. 

It has been reduced quite a bit since then.

--