Anoop Sam John created HBASE-11503:
--------------------------------------
Summary: Inconsistency in CP between HRegion#batchMutate() and
MultiRowMutation()
Key: HBASE-11503
URL: https://issues.apache.org/jira/browse/HBASE-11503
Project: HBase
Issue Type: Bug
Components: Coprocessors, regionserver
Reporter: Anoop Sam John
Assignee: Anoop Sam John
>From users perspective both are almost same except in MultiRowMutation all
>Mutations are applied in one transaction. So the call and behavior of CPs
>during both of these execution should be consistent. Now after HBASE-11423,
>all hooks are getting called.
But in case of HRegion#batchMutate() when a prePut/Delete() call for one
Mutation do a bypass, the actual Mutation we are not applying but will continue
with other Mutations. But in case of MultiRowMutation, when prePut/Delete()
call for one Mutation do a bypass, we just stop calling pre hooks for remaining
Mutations but will end up applying those Mutations to Region.
{code}
public void preProcess(HRegion region, WALEdit walEdit) throws IOException {
RegionCoprocessorHost coprocessorHost = region.getCoprocessorHost();
if (coprocessorHost != null) {
for (Mutation m : mutations) {
if (m instanceof Put) {
if (coprocessorHost.prePut((Put) m, walEdit, m.getDurability())) {
// by pass everything
return;
}
} else if (m instanceof Delete) {
Delete d = (Delete) m;
region.prepareDelete(d);
if (coprocessorHost.preDelete(d, walEdit, d.getDurability())) {
// by pass everything
return;
}
}
}
}
}
{code}
--
This message was sent by Atlassian JIRA
(v6.2#6252)
