I tried to use writethrough, but vms always died with qemu exceptions(qemu 1.5-2.1, sheep 0.8.0) after some minutes/hours/days of work. After that, during one year, I used writeback without any visible problems with data consistency on over 300 images with regular compute nodes crashing(hw/power failures). Just remember to clear object cache directory before starting sheep process at crashed node if you already redeployed vms from that node to others.
-- sheepdog mailing list [email protected] https://lists.wpkg.org/mailman/listinfo/sheepdog
