The RS already returns to the client whether or not it has additional
results to be returned in a subsequent call to scan(), but the ClientScanner
did not use or adhere to this value. Subsequently, this can lead to
bugs around moving to the next region too early. A new method was added
to ClientScanner in the name of testability.
Encapsulate server-state into RegionServerCallable to avoid
modifying parameterization of callable impls.
Signed-off-by: Andrew Purtell <apurtell@apache.org>
Include some basic tests for the method on a testing cluster.
Also update master page to show an alert when balancer is disabled.
Signed-off-by: Enis Soztutar <enis@apache.org>
Summary: Add MemcachedBlockCache
Test Plan: Tested locally with PE and running memcached.
Subscribers: rajesh.nishtala, ndimiduk
Differential Revision: https://reviews.facebook.net/D34635
Summary:
The current behavior of a region move shuts down a region and then starts is up in another regionserver. This causes increased latency and possibly timeouts till the new region's cache is fully warmed up. We can make a region move less disruptive by warming the cache in the destination region server before shutting dow the old region.
See https://issues.apache.org/jira/browse/HBASE-13316
Test Plan:
1. Unit Tests
2. Added test for concurrent moves and warmups
3. Manually tested reads/writes happening with concurrent moves
Subscribers: tedyu
Differential Revision: https://reviews.facebook.net/D35967
Signed-off-by: Elliott Clark <eclark@apache.org>
Summary:
CatalogJanitor#checkDaughterInFs assumes that there are no references
whenever HRegionFileSystem.openRegionFromFileSystem throws IOException.
Well Hadoop and HBase throw IOExceptions whenever someone looks in their general direction.
This patch explicitly checks if the directory exists. If it doesn't then it allows references to be
deleted. All other exceptions cause CatalogJanitor to assume there are references
Test Plan: Unit tests.
Differential Revision: https://reviews.facebook.net/D35829
