[sheepdog-users] Node Recovery
peetee at cheerful.com
Tue Jan 31 13:26:24 CET 2017
A slight confusion on my part in terms of node recovery -
Using Sheepdog 0.8.3 on Ubuntu 16.04 with Corosync and a simple setup.
I have 2 KVM servers set up and 1 cluster.
I have 2 VDI's in the cluster both of 20gb.
Neither VDI is active in the 2 KVM servers (i.e.all virtual machines stopped).
If i reboot (normal shutdown) a server (say node 0) on restart it says "node 0 in recovery" - it than seems to be rebuilding the VDI's on node zero (some 40gb of I/O). Each time I reboot, the same occurs.
This somewhat confuses me. Both VDI disks are identical on each node, no updates/writes have occurred - surely Sheepdog should know nothing has change?
What if I expand this out to a Terabyte of VDI's - taking a node down for maintenance will incur massive delay on restart.
Am I doing something wrong?
It is not re-building. Recheck of I/O activity via IOTOP shows ONLY reads of disk holding cluster (same approximate amount of reads on both nodes)
More information about the sheepdog-users