Remove a dead node from the cluster
Upon researching the subject, it looks like there are only two ways of removing a dead node from the cluster without trying to re-create it first:
vertica_update --remove-host
drop node cascade;
which reportedly may not even be working and the only other way is by contacting Vertica support to use CE.
Am I missing something or is that correct?
Basically, what we're trying to accomplish is to evict an extremely poorly performing node from the cluster in a way that doesn't involve reading terabytes of data from it (K-SAFE=1). And we've so far learned that "remove host from database" triggers a re-balance and so does making the node ephemeral, so they both will most likely read and write to and from the bad node.
0