On Mon, 2008-10-27 at 19:19 +0000, Simon Riggs wrote:
> Single integrated patch with all features.
v3a dev snapshot, applies to head with minor offsets
passes make check and recovery of continuous make installcheck
> New Features
* recovery.conf parameter: max_standby_delay
* startup process waits for, then cancels backends with old snapshots
> Internal changes since last version:
> * various other bug fixes and minor tweaks
> Known Issues
> * one case in "make installcheck" recovery fails an assert check.
> Mulling over a clean way to handle it, since its rare, but fiddly. Tried
> a couple of ways of solving it, neither very satisfactory. I'll sleep on
> it. Not a showstopper, but I don't want to holdoff presenting a snapshot
> because of it.
Fixed, plus 2 other prior bugs.
> Things still in progress
> * row cleaners delay or kill users
> * locking, relcache invalidation (early stages of coding)
> * correct locking around flat file refresh (just dummied out for now)
> * btree vacuuming changes (prototyped, but deferred for now)
> * prepared transactions (regression test disabled currently)
Simon Riggs www.2ndQuadrant.com
PostgreSQL Training, Services and Support
In response to
pgsql-hackers by date
|Next:||From: Tom Lane||Date: 2008-10-30 02:23:36|
|Subject: Re: pre-MED |
|Previous:||From: Simon Riggs||Date: 2008-10-30 02:11:44|
|Subject: Re: BufferAccessStrategy for bulk insert|