zookeeper-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Jose Antonio Mercado <jose_an...@yahoo.com.INVALID>
Subject Zookeeper sync behavior when dataLogDir is missing
Date Thu, 06 Dec 2018 21:17:25 GMT
We experienced the following issue while doing a rolling restart of our five-server Zookeeper
cluster. While server id 5 was rebooted the filesystem with the contents of Zookeeper's dataDir
was reformatted by mistake, however the contents of dataLogDir was left intact as it is stored
in a different filesystem. When Zookeeper started it didn't find the dataDir so it crated
a new one and started a FastLeaderElection. The problem was server id 5 did not get a snapshot
from the leader but a diff instead. So, server 5 ended up with an empty znode space while
the other servers had all znodes intact.
Does anyone know what is the expected behavior if the contents of the dataDir go missing but
all transaction logs in datalogDir are still present when Zookeeper is restarted? 

  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message