When creating a cluster of 3 or 5 members, and then stopping the master, it seems that one of the members fails to rejoin the new master; and instead dies with a Python error:
Thoughts on what might be going on? My Python isn't strong; so can't easily spot what data might be flowing down into etcd.py that causes AttributeError: 'NoneType' object has no attribute 'strip'
This issue did not happen when we then stopped the new master.
That is, when we had 5 members, if we stop the master, we lose the master and one of the replicas and the cluster size is 3. When we stop the new master, we only lose the master, and the cluster size is 2.
When we start with a cluster of 3 members, if we stop the master we lose the master and one of the replicas; and the final cluster size is 1.
When creating a cluster of 3 or 5 members, and then stopping the master, it seems that one of the members fails to rejoin the new master; and instead dies with a Python error:
https://gist.github.com/drnic/b6e4fbd9940d8694dded
Thoughts on what might be going on? My Python isn't strong; so can't easily spot what data might be flowing down into etcd.py that causes
AttributeError: 'NoneType' object has no attribute 'strip'
This issue did not happen when we then stopped the new master.
That is, when we had 5 members, if we stop the master, we lose the master and one of the replicas and the cluster size is 3. When we stop the new master, we only lose the master, and the cluster size is 2.
When we start with a cluster of 3 members, if we stop the master we lose the master and one of the replicas; and the final cluster size is 1.