11 апреля 2014 г. 16:20:32 CEST, John McEntee <jmcen...@stirling-dynamics.com> пишет: >Due to pushing each disk fairly hard , if the zpool gets more >than 80% full the way free space is allocated changes and we have a >severe performance hit. This makes (generally one) VMWare host >virtually unusable until I free up some space. Does anyone know if this >behaviour changes due to the result of zpool or zfs. i.e. which one >should I carefully monitor?
In short, this is a known issue with zfs (search the archives). IIRC, the more full a pool is, the longer it takes the write allocator to find sufficiently long free space stretches (and also this causes more fragmentation which impacts subsequent reads at least from spinning rust). And the effect comes into action quite sharply, although the exact number depends on pre-history of the particular pool - ranging from 70% full to well over 90%. Generally the solution is to either free up some space or to grow the pool (and maybe rebalance existing data if you added vdevs). Hth, //Jim -- Typos courtesy of K-9 Mail on my Samsung Android _______________________________________________ OpenIndiana-discuss mailing list OpenIndiana-discuss@openindiana.org http://openindiana.org/mailman/listinfo/openindiana-discuss
