[
https://issues.apache.org/jira/browse/CURATOR-76?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Jay Bae updated CURATOR-76:
---------------------------
Description: We are having serious data corruption issue when we are
rolling restart of zookeeper servers due to one application which is using
ChildReaper recipe. I am not sure its root cause but my theory is, when the
multiple instances are running ChildReaper recipe, they would conflict each
other among checking exist and deleting paths. This conflict can cause data
corruption. We observed all servers died due to corrupted data and we had to
manually copy log/snapshot data and restart them. (was: We are having serious
data corruption issue when we are rolling restart of zookeeper servers due to
one application which is using ChildReaper recipe. I am not sure its root cause
but my theory is, when the multiple instances are running ChildReaper recipe,
they would conflict each other among checking exist and deleting paths. This
conflict can cause data corruption. We observed all servers died due to
corrupted data and we had to manually copy log/snapshot data and restart them.
Also, it wouldn't be enough checking simply whether the zknode is empty. It
would be better if ChildReaper is checking the node is empty and it's not
modified for the amount of time.)
Issue Type: Improvement (was: Bug)
Summary: Adding leader selection ChildReaper recipe (was: Adding
leader selection and TTL feature in ChildReaper recipe)
> Adding leader selection ChildReaper recipe
> ------------------------------------------
>
> Key: CURATOR-76
> URL: https://issues.apache.org/jira/browse/CURATOR-76
> Project: Apache Curator
> Issue Type: Improvement
> Components: Recipes
> Reporter: Jay Bae
>
> We are having serious data corruption issue when we are rolling restart of
> zookeeper servers due to one application which is using ChildReaper recipe. I
> am not sure its root cause but my theory is, when the multiple instances are
> running ChildReaper recipe, they would conflict each other among checking
> exist and deleting paths. This conflict can cause data corruption. We
> observed all servers died due to corrupted data and we had to manually copy
> log/snapshot data and restart them.
--
This message was sent by Atlassian JIRA
(v6.1#6144)
