Any alternatives to rolling cluster back to LGE?
After an unclean database shutdown oftentimes the database would not start unless we roll it back to the last good epoch. That implies a certain loss of data. I've been asked what alternatives are there? Is there any situation imaginable when we can avoid rolling the cluster back and yet get it up without losing any data? How would we go about it?
Say, on a 3-node cluster, two nodes get updated and one doesn't, so now it is lagging behind. Rolling back to LGE would essentially move the two updated nodes back, correct? How about doing it the other way around - moving the lagging node forward? Or did I get it all wrong?
0
Answers
In order to provide the options, we would need to review your cluster. please open a support case.
That is not based on a specific situation. More like a general question. I've run into an ASR required situation quite a lot on different systems and I just routinely use Roll back to LGE, assuming some data loss is inevitable and the rollback will keep it to the minimum. But really in an ASR required situation on a K-SAFE1 cluster are there any other recovery options to consider ever (because rolling back to LGE would work but would lose more data than is absolutely necessary)?