We are doing a test simulating disk fail. The test step is described as follows:
1. start sheep on each node like this: sheep -n /home/admin/sheepdogmetadata/,/disk1/sheepdogstoredata/,/disk2/sheepdogstoredata/,/disk3/sheepdogstoredata/,/disk4/sheepdogstoredata/,/disk5/sheepdogstoredata/,/disk6/sheepdogstoredata/,/disk7/sheepdogstoredata/,/disk8/sheepdogstoredata/,/disk9/sheepdogstoredata/ -b 0.0.0.0 -y *.*.*.* -c zookeeper:*.*.*.*:2181 2. umount a disk on a node: sudo umount /disk5 3. then sheep.log on this node output: Jun 10 20:55:37 ERROR [io 122878] err_to_sderr(96) /disk5/sheepdogstoredata corrupted Jun 10 20:55:37 ERROR [io 122878] err_to_sderr(96) /disk5/sheepdogstoredata corrupted Jun 10 20:55:37 ERROR [io 122871] err_to_sderr(96) /disk5/sheepdogstoredata corrupted Jun 10 20:55:37 ERROR [io 122702] err_to_sderr(96) /disk5/sheepdogstoredata corrupted The log is growing fast, and may fill the log disk. -------------------------------------------------------------------------------------------------------------------------------- Should sheep panic and exit when the store path does not exist? If sheep on this node does not exit, all vms on other nodes want to read or write on this store path may hang. Do you agree with this? -- Xu Fang
-- sheepdog mailing list [email protected] http://lists.wpkg.org/mailman/listinfo/sheepdog
