Hi @weibin.wu
Can you please share here the configuration of your cluster?
In particular, I want to check how many availability zones (AZ) are you using. I guess you are using 2 availability zones (b-1 and b-2) the third zone (a-1) is used for the tie breaker. If this is the case then the behavior you are experiencing in both cases is expected.
During the first attempt, since tiebreaker is hosted on a-1 then our algorithm does not allow ES data to be moved to the same AZ with the tiebreaker.
During the second attempt, we have disabled completely the b-1 zone and since a-1 cannot be used (for the same reason as explained above) the algorithm tries to use the only available/allowed zone which is b-1 and it fails because it can't be used.
Apache, Apache Lucene, Apache Hadoop, Hadoop, HDFS and the yellow elephant
logo are trademarks of the
Apache Software Foundation
in the United States and/or other countries.