> > Deleting the dedup'ed data won't work better, > since > > ZFS will have to process it quite the same way as > if > > you're destroying a ZFS volume. > > This was my only dedup'd dataset, so I have no other > experience deleting them. Do you think that, had > there been no data, my "zfs destroy" would have taken > 3 days? I get that the rm -rf might have taken 3 > days, but hopefully my system wouldn't have gone all > freaky during that?
No, I was going to say that manually deleting the deduped files would have taken as long as the zfs destroy had taken, but your system would likely have behaved pretty good during that time. > > >On the other hand, there're quite a number of > calucations that'll give you a good > >guess about how much RAM you will need for a > specific amount of dedup'ed data. > > I think an extremely safe calc would say 3GB RAM per > 1TB of pre-dedup'd storage. Our data was 2TB before > dedup. We have 16GB of RAM in this unit so we > shouldn't have been having a RAM issue. We had > plenty of free RAM during the whole process. Afaik, this is extremely dependent on the "level" of deduped objects. I had a box that had 32 GB of RAM and that went nuts with dedup when the the volume exceeded 8 TB - but that was on snv_134. I have never used dedup afterwards. budy -- This message posted from opensolaris.org _______________________________________________ opensolaris-discuss mailing list opensolaris-discuss@opensolaris.org
