This patchset fixes the problem reported in the sheepdog-user list. http://lists.wpkg.org/pipermail/sheepdog-users/2012-July/000200.html
When there are many stale objects and those sha1 objects are not in the farm backend store, sheep blocks I/Os long time. On my environment, when there are 50 GB data to be moved to the farm backend, sheep blocks I/Os 15 minitues (not seconds). It is because sheep do the work in farm_end_recover(), but the function is called in the main thread. This patchset moves the function to the worker thread, and fix relevant problems. MORITA Kazutaka (2): farm: call farm_end_recover in worker thread farm: recover object from omap when trunk is not found sheep/farm/farm.c | 15 +++++++++++++++ sheep/farm/farm.h | 1 + sheep/farm/trunk.c | 39 +++++++++++++++++++++++++++++++++++++-- sheep/recovery.c | 28 +++++++++++++++++++++++----- 4 files changed, 76 insertions(+), 7 deletions(-) -- 1.7.2.5 -- sheepdog mailing list [email protected] http://lists.wpkg.org/mailman/listinfo/sheepdog
