Обсуждение: important Re: [HACKERS] Open 6.5 items
Hiroshi wrote: > Ole Gjerde who provided the patch for current implementation of > mdtruncate() sayz. > "First, please reverse my patch to mdtruncate() in md.c as soon as > possible. It does not work properly in some cases." > > I also recommend to reverse his patch to mdtruncate(). > > Though we could not shrink segmented relations by old implementation > the result by vacuum would never be inconsistent(?). > > I think we don't have enough time to fix this. > If there is no fix for vacuum, I suggest to change the filesize before splitting back to just below 2 Gb (2Gb - 8k). Else vacuum will only work for tables up to 1 Gb, and it did work up to 2 Gb before. I am the one who suggested 1 Gb, so I had my eye on this issue. I still think 1 Gb is good for various reasons, but only if vacuum works. Andreas
ZEUGSWETTER Andreas IZ5 wrote: > > Hiroshi wrote: > > Ole Gjerde who provided the patch for current implementation of > > mdtruncate() sayz. > > "First, please reverse my patch to mdtruncate() in md.c as soon as > > possible. It does not work properly in some cases." > > > > I also recommend to reverse his patch to mdtruncate(). > > > > Though we could not shrink segmented relations by old implementation > > the result by vacuum would never be inconsistent(?). > > > > I think we don't have enough time to fix this. > > > If there is no fix for vacuum, I suggest to change the filesize before > splitting > back to just below 2 Gb (2Gb - 8k). Else vacuum will only work for tables > up to 1 Gb, and it did work up to 2 Gb before. > > I am the one who suggested 1 Gb, so I had my eye on this issue. > I still think 1 Gb is good for various reasons, but only if vacuum works. Is this issue addressed by last mdtruncate() changes? Vadim
> > I am the one who suggested 1 Gb, so I had my eye on this issue. > > I still think 1 Gb is good for various reasons, but only if vacuum works. > > Is this issue addressed by last mdtruncate() changes? > I think it is fixed. -- Bruce Momjian | http://www.op.net/~candle maillist@candle.pha.pa.us | (610) 853-3000+ If your life is a hard drive, | 830 Blythe Avenue + Christ can be your backup. | Drexel Hill, Pennsylvania19026
> Hiroshi wrote: > > Ole Gjerde who provided the patch for current implementation of > > mdtruncate() sayz. > > "First, please reverse my patch to mdtruncate() in md.c as soon as > > possible. It does not work properly in some cases." > > > > I also recommend to reverse his patch to mdtruncate(). > > > > Though we could not shrink segmented relations by old implementation > > the result by vacuum would never be inconsistent(?). > > > > I think we don't have enough time to fix this. > > > If there is no fix for vacuum, I suggest to change the filesize before > splitting > back to just below 2 Gb (2Gb - 8k). Else vacuum will only work for tables > up to 1 Gb, and it did work up to 2 Gb before. > > I am the one who suggested 1 Gb, so I had my eye on this issue. > I still think 1 Gb is good for various reasons, but only if vacuum works. This is where we dropped the ball. We should have made this recommended change before 6.5. -- Bruce Momjian | http://www.op.net/~candle maillist@candle.pha.pa.us | (610) 853-3000+ If your life is a hard drive, | 830 Blythe Avenue + Christ can be your backup. | Drexel Hill, Pennsylvania19026