On Thu, 17 Jan 2013, Christopher Kunz wrote: > Hi, > > if run during a user-issued reweight (i.e. "ceph osd crush reweight x > y"), "ceph osd tree" shows the target weight of an OSD. Is there a way > to see the *current* weight of the OSD? > > We would like to be able to approximate the amount of rollback necessary > per OSD if we need to cancel a larger reweight at some point. This seems > impossible, though, without knowing the current weight. > > Any hints how to do this? There are two things you have: the previous weight (before you changed it, if you remember), and what you see from 'df'. The latter is found in the tail end of the 'ceph pg dump' output. The thing to keep in mind, though, is that rolling back doesn't generally let you capture partial rebalance progress. If you change the weight of a node from 1 to .8, wait for about half of the data to move, and then change it to .9, half will be remapped back to the original location, but it is a different random subset of the data.. which mean that 1/4 will have been moved and be in the right place, 1/4 will have moved but have to move back, 1/4 will not have moved but still need to move, and 1/4 will have not moved and not need to. So in the end, that "half-way point" is similar to the point of no return: it's just as much work to go back as it is to finish. If you are going to correct mid-stream, I would try to do so quickly, and probably just go back to the original weight instead of something in-between. sage > > Regards, > > --ck > > -- > filoo GmbH > Dr. Christopher Kunz > > E-Mail: chris@xxxxxxxx > Tel.: (+49) 0 52 41 8 67 30 -18 > Fax: (+49) 0 52 41 / 8 67 30 -20 > > Please sign & encrypt mail wherever possible, my key: > C882 8ED1 7DD1 9011 C088 EA50 5CFA 2EEB 397A CAC1 > > Moltkestra?e 25a > 33330 G?tersloh > > HRB4355, AG G?tersloh > Gesch?ftsf?hrer: S.Grewing, J.Rehp?hler, Dr. C.Kunz > > Filoo im Web: http://www.filoo.de/ > Folgen Sie uns auf Twitter: http://twitter.com/filoogmbh > Werden Sie unser Fan auf Facebook: http://facebook.com/filoogmbh > -- > To unsubscribe from this list: send the line "unsubscribe ceph-devel" in > the body of a message to majordomo@xxxxxxxxxxxxxxx > More majordomo info at http://vger.kernel.org/majordomo-info.html > > -- To unsubscribe from this list: send the line "unsubscribe ceph-devel" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
