[4/4] kudu git commit: docs: updates to kudu_impala_integration.adoc

2016-09-01 Thread adar
docs: updates to kudu_impala_integration.adoc Note that this doc has diverged pretty substantially from its downstream Cloudera counterpart, which is far more featureful. Change-Id: I02a91f1c33be3f0d6fbffaef5e03832e21b6db70 Reviewed-on: http://gerrit.cloudera.org:8080/4138 Tested-by: Kudu Jenkins

[3/4] kudu git commit: cache: dynamically determine the number of cache shards

2016-09-01 Thread adar
cache: dynamically determine the number of cache shards I noticed in a heavily cache-bound workload on a 48-core machine that there was a lot of contention on the cache shard spin locks. Manually bumping the number of cache shards to 64 instead of the default 16 seemed to reduce the amount of cont

[1/4] kudu git commit: KUDU-1495. Maintenance manager should not schedule new ops during unregister

2016-09-01 Thread adar
Repository: kudu Updated Branches: refs/heads/master 93be1310d -> d13ac79a4 KUDU-1495. Maintenance manager should not schedule new ops during unregister This fixes a bug where deleting a tablet could block for an extended time on compactions being scheduled if the maintenance manager was confi

[2/4] kudu git commit: KUDU-1557. maintenance_manager: return most recently completed ops first

2016-09-01 Thread adar
KUDU-1557. maintenance_manager: return most recently completed ops first This changes the order of maintenance manager operations in the maintenance manager status dump so that the most recently completed operation is first in the list. Change-Id: I7783bc8cee1beb0f7accaedbd46fb1e3a49168cc Reviewe

[1/4] kudu git commit: maintenance_manager: simplify test

2016-09-01 Thread todd
Repository: kudu Updated Branches: refs/heads/master 000791109 -> 93be1310d maintenance_manager: simplify test This simplifies maintenance_manager-test a bit to avoid a condition variable, and instead just poll from the main thread. This reduces the complexity to understand the test a bit, and

[3/4] kudu git commit: KUDU-456 Implement AUTO_FLUSH_BACKGROUND flush mode

2016-09-01 Thread todd
http://git-wip-us.apache.org/repos/asf/kudu/blob/93be1310/src/kudu/client/session-internal.cc -- diff --git a/src/kudu/client/session-internal.cc b/src/kudu/client/session-internal.cc index 552b507..8da7679 100644 --- a/src/kudu/c

[2/4] kudu git commit: tserver: sort /tablets list by table name first

2016-09-01 Thread todd
tserver: sort /tablets list by table name first I noticed when I'm looking at /tablets I usually am most interested in looking at tablets from a particular table. So, sorting by tablet ID is not that useful. This patch changes to sorting by table name first, and then within each table sorting by t

[4/4] kudu git commit: KUDU-456 Implement AUTO_FLUSH_BACKGROUND flush mode

2016-09-01 Thread todd
KUDU-456 Implement AUTO_FLUSH_BACKGROUND flush mode Implemented AUTO_FLUSH_BACKGROUND for the Kudu C++ client library. In AUTO_FLUSH_BACKGROUND mode, the KuduSession::Apply() method blocks if total amount of data for pending operations reaches the buffer size limit. The limit on the buffer size c

kudu git commit: docs: update installation with new OS support

2016-09-01 Thread todd
Repository: kudu Updated Branches: refs/heads/master 680cec6bb -> 000791109 docs: update installation with new OS support At the time of writing, a couple things are broken: 1. The SLES 12, Jessie, and Xenial Cloudera repo files contain unsubstituted template variables. 2. The SLES 12 Cloud

kudu git commit: ts_tablet_manager: set status message after failed tablet copy

2016-09-01 Thread adar
Repository: kudu Updated Branches: refs/heads/master e1f54cbe7 -> 680cec6bb ts_tablet_manager: set status message after failed tablet copy I noticed in a test cluster that a tablet copy had failed, and ksck was reporting the state as: b314b10216d74c07ab9da974671137db (ve0136.halxg.cloudera.

[1/2] kudu git commit: ksck: add --checksum_scan_concurrency to optional flags

2016-09-01 Thread adar
Repository: kudu Updated Branches: refs/heads/master 075c42cb7 -> e1f54cbe7 ksck: add --checksum_scan_concurrency to optional flags Change-Id: Iaaa76215e0c30860eebd99db0548b3b37dd2e60e Reviewed-on: http://gerrit.cloudera.org:8080/4183 Tested-by: Kudu Jenkins Reviewed-by: Dan Burkert Project

[2/2] kudu git commit: tablet_peer: clean up TabletStatusListener interface

2016-09-01 Thread adar
tablet_peer: clean up TabletStatusListener interface This class duplicated a lot of TabletPeer info for no particular reason. Its main purpose was to provide a clean callback for reporting status back to the TabletPeer, but over time it grew various unnecessary warts. Change-Id: I06233b88a8385902

[2/2] kudu git commit: transaction_tracker: back-off when logging in-flight transactions

2016-09-01 Thread todd
transaction_tracker: back-off when logging in-flight transactions On a test cluster pushed into overload, I ended up with thousands of in-flight transactions which were only being drained at the rate of one or two per second. Dumping the entire list of transactions once a second ended up flooding

[1/2] kudu git commit: tracing: gzip the trace JSON ajax response

2016-09-01 Thread todd
Repository: kudu Updated Branches: refs/heads/master 4fd0572ee -> 075c42cb7 tracing: gzip the trace JSON ajax response This enables gzipping of the actual trace which is sent back over HTTP when using the tracing web page. I noticed that the traces can be quite slow to receive, but given they

kudu git commit: [java-client] Add ScanToken.stringifySerializedToken

2016-09-01 Thread danburkert
Repository: kudu Updated Branches: refs/heads/master 1f660ee9f -> 4fd0572ee [java-client] Add ScanToken.stringifySerializedToken stringifySerializedToken takes a serialized scan token, and returns a String suitable for debug printing. The string contains information sufficient to determine whi