I hit an assert in some of my manual fiddling with the tiering agent where a promote, copy-from, or flush operation would hit -546> 2014-01-20 14:44:55.506594 7f0a467fc700 -1 osd/ReplicatedPG.cc: In function 'void ReplicatedPG::finish_ctx(ReplicatedPG::OpContext*, int)' thread 7f0a467fc700 time 2014-01-20 14:44:55.466752 osd/ReplicatedPG.cc: 4737: FAILED assert(soid < scrubber.start || soid >= scrubber.end) The problem is that we normally block ops that hit the current scrub chunk in do_op at the very top level, but all of these ops are initiating writes at lower levels after doing some other slow/blocking work. They are smart enough to take and block on the rwlock stuff, but I'm worried that the waiting_for_active queue is too coarse for this. My thought is to add a std::map of obc's blocked on scrub and wake them up when the scrub chunk completes. Any better ideas? Incidentally, we need to make the thrashosds task smart enough to pause thrashing long enough for scrubs to happen so that these issues are covered in the normal test suite... sage -- 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
