Looking at the files AZURE_PING creates in the storage container, it seems that when a cluster member leaves the cluster (a deployment is undeployed, or the server is shut down), it's record is not purged from the AZURE_PING file. According to the FILE_PING design document, records of cluster members that have left should be removed.

When the last cluster member is properly shut down, the AZURE_PING file is correctly removed.

So far, I haven't observed if this has any negative impact on the cluster.

I'm attaching three files (second and third have '(X)' at the end of their names). The first one shows an entry for a two node cluster. The second one shows an entry for the same cluster once the first node was shut down (sst-rjanik-fd4). The third file shows the same cluster once both nodes have been sequentially shut down and started again (four entries, but only two are valid).

Assigning to rhn-engineering-rhusar since he's the author of AZURE_PING. Of course, feel free to reassign as you see fit.

Not setting affects docs, because this doesn't seem to affect anything.