[
https://issues.apache.org/jira/browse/HBASE-12575?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Andrew Purtell updated HBASE-12575:
-----------------------------------
Attachment: HBASE-12575.patch
Ok, this is ready. Let's just check table coprocessor class availability and
call it a day as this is the reported problem.
We could do just a bit more and check if the store engine class is loadable,
but that isn't very useful. A store engine implementation determines what
flusher, compactor, or compaction policy class to use. The choices are
unknowable without an instance. We _could_ create a test StoreEngine instance
in the master, but we'd need to create a real test Store first - components
like the compactor want to look at a Store object for configuration in their
constructor - and so I think that gets out of hand.
> Sanity check table coprocessor classes are loadable
> ---------------------------------------------------
>
> Key: HBASE-12575
> URL: https://issues.apache.org/jira/browse/HBASE-12575
> Project: HBase
> Issue Type: Sub-task
> Reporter: Enis Soztutar
> Assignee: Andrew Purtell
> Fix For: 1.0.0, 2.0.0, 0.98.9
>
> Attachments: HBASE-12575.patch, HBASE-12575.patch, HBASE-12575.patch,
> HBASE-12575.patch
>
>
> We load coprocessors and other classes from configuration. In case of a typo
> in the class name (or deployment problem) a create table / alter table with
> wrong class name brings down the whole cluster.
> Master already does sanity check for compression and region split policy
> classes introduced in HBASE-10591. We should extend that to check some other
> common cases as well.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
