Hello Loic, sounds good but my initial question was if this shouldn't be integrated in ceph-deploy - so when you add OSDs it also does the correct reweight? Greets, Stefan Am 14.05.2017 um 19:46 schrieb Loic Dachary: > Hi Stefan, > > A new python-crush[1] subcommand will be available next week that you could use to rebalance your clusters. You give it a crushmap and it optimizes the weights to fix the uneven distribution. It can produce a series of crushmaps, each with a small modification so that you can gradually improve the situation and better control how many PGs are moving. > > Would that be useful for the clusters you have ? > > Cheers > > [1] http://crush.readthedocs.io/ > > On 05/02/2017 09:32 AM, Loic Dachary wrote: >> >> >> On 05/02/2017 07:43 AM, Stefan Priebe - Profihost AG wrote: >>> Hi Loic, >>> >>> yes i didn't changed them to straw2 as i didn't saw any difference. I >>> switched to straw2 now but it didn't change anything at all. >> >> straw vs straw2 is not responsible for the uneven distribution you're seeing. I meant to say the optimization only works on straw2 buckets, it is not implemented for straw buckets. >> >>> If i use those weights manuall i've to adjust them on every crush change >>> on the cluster? That's something i don't really like to do. >> >> This is not practical indeed :-) I'm hoping python-crush can automate that. >> >> Cheers >> >>> Greets, >>> Stefan >>> >>> Am 02.05.2017 um 01:12 schrieb Loic Dachary: >>>> It is working, with straw2 (your cluster still is using straw). >>>> >>>> For instance for one host it goes from: >>>> >>>> ~expected~ ~objects~ ~over/under used %~ ~delta~ ~delta%~ >>>> ~name~ >>>> osd.24 149 159 6.65 10.0 6.71 >>>> osd.29 149 159 6.65 10.0 6.71 >>>> osd.0 69 77 11.04 8.0 11.59 >>>> osd.2 69 69 -0.50 0.0 0.00 >>>> osd.42 149 148 -0.73 -1.0 -0.67 >>>> osd.1 69 62 -10.59 -7.0 -10.14 >>>> osd.23 69 62 -10.59 -7.0 -10.14 >>>> osd.36 149 132 -11.46 -17.0 -11.41 >>>> >>>> to >>>> >>>> ~expected~ ~objects~ ~over/under used %~ ~delta~ ~delta%~ >>>> ~name~ >>>> osd.0 69 69 -0.50 0.0 0.00 >>>> osd.23 69 69 -0.50 0.0 0.00 >>>> osd.24 149 149 -0.06 0.0 0.00 >>>> osd.29 149 149 -0.06 0.0 0.00 >>>> osd.36 149 149 -0.06 0.0 0.00 >>>> osd.1 69 68 -1.94 -1.0 -1.45 >>>> osd.2 69 68 -1.94 -1.0 -1.45 >>>> osd.42 149 147 -1.40 -2.0 -1.34 >>>> >>>> By changing the weights to >>>> >>>> [0.6609248140022604, 0.9148542821020436, 0.8174711575190294, 0.8870680217468655, 1.6031393139865695, 1.5871079208467038, 1.8784764188501162, 1.7308530904776616] >>>> >>>> And you could set these weights on the crushmap, there would be no need for backporting. >>>> >>>> >>>> On 05/01/2017 08:06 PM, Stefan Priebe - Profihost AG wrote: >>>>> Am 01.05.2017 um 19:47 schrieb Loic Dachary: >>>>>> Hi Stefan, >>>>>> >>>>>> On 05/01/2017 07:15 PM, Stefan Priebe - Profihost AG wrote: >>>>>>> That sounds amazing! Is there any chance this will be backported to jewel? >>>>>> >>>>>> There should be ways to make that work with kraken and jewel. It may not even require a backport. If you know of a cluster with an uneven distribution, it would be great if you could send the crushmap so that I can test the algorithm. I'm still not sure this is the right solution and it would help confirm that. >>>>> >>>>> I've lots of them ;-) >>>>> >>>>> Will sent you one via private e-mail in some minutes. >>>>> >>>>> Greets, >>>>> Stefan >>>>> >>>>>> Cheers >>>>>> >>>>>>> >>>>>>> Greets, >>>>>>> Stefan >>>>>>> >>>>>>> Am 30.04.2017 um 16:15 schrieb Loic Dachary: >>>>>>>> Hi, >>>>>>>> >>>>>>>> Ideally CRUSH distributes PGs evenly on OSDs so that they all fill in >>>>>>>> the same proportion. If an OSD is 75% full, it is expected that all >>>>>>>> other OSDs are also 75% full. >>>>>>>> >>>>>>>> In reality the distribution is even only when more than 100,000 PGs >>>>>>>> are distributed in a pool of size 1 (i.e. no replication). >>>>>>>> >>>>>>>> In small clusters there are a few thousands PGs and it is not enough >>>>>>>> to get an even distribution. Running the following with >>>>>>>> python-crush[1], shows a 15% difference when distributing 1,000 PGs on >>>>>>>> 6 devices. Only with 1,000,000 PGs does the difference drop under 1%. >>>>>>>> >>>>>>>> for PGs in 1000 10000 100000 1000000 ; do >>>>>>>> crush analyze --replication-count 1 \ >>>>>>>> --type device \ >>>>>>>> --values-count $PGs \ >>>>>>>> --rule data \ >>>>>>>> --crushmap tests/sample-crushmap.json >>>>>>>> done >>>>>>>> >>>>>>>> In larger clusters, even though a greater number of PGs are >>>>>>>> distributed, there are at most a few dozens devices per host and the >>>>>>>> problem remains. On a machine with 24 OSDs each expected to handle a >>>>>>>> few hundred PGs, a total of a few thousands PGs are distributed which >>>>>>>> is not enough to get an even distribution. >>>>>>>> >>>>>>>> There is a secondary reason for the distribution to be uneven, when >>>>>>>> there is more than one replica. The second replica must be on a >>>>>>>> different device than the first replica. This conditional probability >>>>>>>> is not taken into account by CRUSH and would create an uneven >>>>>>>> distribution if more than 10,000 PGs were distributed per OSD[2]. But >>>>>>>> a given OSD can only handle a few hundred PGs and this conditional >>>>>>>> probability bias is dominated by the uneven distribution caused by the >>>>>>>> low number of PGs. >>>>>>>> >>>>>>>> The uneven CRUSH distributions are always caused by a low number of >>>>>>>> samples, even in large clusters. Since this noise (i.e. the difference >>>>>>>> between the desired distribution and the actual distribution) is >>>>>>>> random, it cannot be fixed by optimizations methods. The >>>>>>>> Nedler-Mead[3] simplex converges to a local minimum that is far from >>>>>>>> the optimal minimum in many cases. Broyden–Fletcher–Goldfarb–Shanno[4] >>>>>>>> fails to find a gradient that would allow it to converge faster. And >>>>>>>> even if it did, the local minimum found would be as often wrong as >>>>>>>> with Nedler-Mead, only it would go faster. A least mean squares >>>>>>>> filter[5] is equally unable to suppress the noise created by the >>>>>>>> uneven distribution because no coefficients can model a random noise. >>>>>>>> >>>>>>>> With that in mind, I implemented a simple optimization algorithm[6] >>>>>>>> which was first suggested by Thierry Delamare a few weeks ago. It goes >>>>>>>> like this: >>>>>>>> >>>>>>>> - Distribute the desired number of PGs[7] >>>>>>>> - Subtract 1% of the weight of the OSD that is the most over used >>>>>>>> - Add the subtracted weight to the OSD that is the most under used >>>>>>>> - Repeat until the Kullback–Leibler divergence[8] is small enough >>>>>>>> >>>>>>>> Quoting Adam Kupczyk, this works because: >>>>>>>> >>>>>>>> "...CRUSH is not random proces at all, it behaves in numerically >>>>>>>> stable way. Specifically, if we increase weight on one node, we >>>>>>>> will get more PGs on this node and less on every other node: >>>>>>>> CRUSH([10.1, 10, 10, 5, 5]) -> [146(+3), 152, 156(-2), 70(-1), 76]" >>>>>>>> >>>>>>>> A nice side effect of this optimization algorithm is that it does not >>>>>>>> change the weight of the bucket containing the items being >>>>>>>> optimized. It is local to a bucket with no influence on the other >>>>>>>> parts of the crushmap (modulo the conditional probability bias). >>>>>>>> >>>>>>>> In all tests the situation improves at least by an order of >>>>>>>> magnitude. For instance when there is a 30% difference between two >>>>>>>> OSDs, it is down to less than 3% after optimization. >>>>>>>> >>>>>>>> The tests for the optimization method can be run with >>>>>>>> >>>>>>>> git clone -b wip-fix-2 http://libcrush.org/dachary/python-crush.git >>>>>>>> tox -e py27 -- -s -vv -k test_fix tests/test_analyze.py >>>>>>>> >>>>>>>> If anyone think of a reason why this algorithm won't work in some >>>>>>>> cases, please speak up :-) >>>>>>>> >>>>>>>> Cheers >>>>>>>> >>>>>>>> [1] python-crush http://crush.readthedocs.io/ >>>>>>>> [2] crush multipick anomaly http://marc.info/?l=ceph-devel&m=148539995928656&w=2 >>>>>>>> [3] Nedler-Mead https://en.wikipedia.org/wiki/Nelder%E2%80%93Mead_method >>>>>>>> [4] L-BFGS-B https://docs.scipy.org/doc/scipy-0.18.1/reference/optimize.minimize-lbfgsb.html#optimize-minimize-lbfgsb >>>>>>>> [5] Least mean squares filter https://en.wikipedia.org/wiki/Least_mean_squares_filter >>>>>>>> [6] http://libcrush.org/dachary/python-crush/blob/c6af9bbcbef7123af84ee4d75d63dd1b967213a2/tests/test_analyze.py#L39 >>>>>>>> [7] Predicting Ceph PG placement http://dachary.org/?p=4020 >>>>>>>> [8] Kullback–Leibler divergence https://en.wikipedia.org/wiki/Kullback%E2%80%93Leibler_divergence >>>>>>>> >>>>>>> -- >>>>>>> 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 >>>>> >>>> >>> >> > -- 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
