Hi Milko, Your setup looks correct. We use Capistrano to launch Hypertable. Our default Capistrano config file launches the servers in the same way as you have listed. The random-wait.sh is also called (in our Capistrano config) before launching the servers on all the slaves. In certain environments with 100+ slave nodes, we we were seeing a thundering herd problem with Hyperspace when they all came up at the same time. This script will wait some random interval between 0 and 5 seconds which staggers the startup times for each slave.
- Doug On Thu, Jul 30, 2009 at 1:07 PM, <[email protected]> wrote: > > Hi Doug, > > I've just installed the new Hypertable release. All tests have been passed > successfully. Tomorrow I'll test the new release with kfs (on 4 nodes). > > Thanks for your efforts to improve the quality of Hypertable DBMS!!! > > Please let me know what is the function of random-wait.sh? When it is run? > > With KFS the following processes are run: > > ON MASTER NODE: > =============== > start-dfsbroker.sh kfs > start-hyperspace.sh > start-master.sh > > ON SLAVE NODES: > =============== > start-dfsbroker.sh kfs > start-rangeserver.sh > start-thriftbroker.sh > > Please let me know if this is correct? > > Or I should run only "start-all-servers.sh" on master node? > > Regards, > > Milko > > ************************************************** > Assoc. Prof. Milko Todorov Marinov, PhD > Department of Computer Systems & Technologies > University of Rousse > 8 Studentska Str, 7017 Rousse, BULGARIA > Tel. (+359 82) 888 356, Fax. (+359 82) 845 708 > Cell phone: (+359 88) 999 6793 > E-mail: [email protected] > ************************************************* > > On Thu, July 30, 2009 4:36 am, Doug Judd wrote: > > Hypertable patch release 0.9.2.5 has been released and is now available > > for download at http://www.hypertable.org/download.html. > > > > NOTE: you may have to hit refresh in your browser to get the correct > > download > > > > Version 0.9.2.5: > > (2009-07-29) > > > > > > Fixed memory leak in query path > > Implemented purgeable CellStore indexes and bloom filters > > Added mutator options to periodically flush buffers > > Fixed bug in CommitLog that was causing some fragments to get dropped > > incorrectly. Fixed bug in CommitLogReader caused by fragment queue > iterator > > invalidation when log fragments are pushed to the back of the deque. > > Rewrote Cell Cache/Store Scanner > > Added MaintenancePrioritizerLowMemory used in low memory condition > > Implemented mutator no flag sync feature and added regression test. > > Fixed performance regression in IntervalScanner > > [Issue 313] Added temporary fix and instrumentation for debugging. > > Updated ruby tests for mutator API changes (The rescue block was masking > > exceptions) Fixed load-exception test > > Setting default commit log revision to TIMESTAMP_MIN instead of 0. > > Fixed commit-log-gc test > > Renamed ClientException::what to message for compatibility with most > > languages Added hql_*2 APIs for returning cells as arrays in HQL queries > > (To > > support find_by_sql in HyperRecord) Allow 'row' and 'column' for LDI > > hyperformat header Regenerated files with updated thrift > > Added --query-delay=<millis> option to load_generator > > Got rid of some java warnings > > Fixed KosmosBroker::pread to not modify current offset > > Thrift related changes for periodic mutator flush > > Rename a test label to follow the convention of the rest of the tests > > Restart test servers automatically every time when make *tests > > Added script to dump backtraces of running/hanging servers. > > Allow VALGRIND to be defined as environment variable in start scripts. > > Added missing php files for the no_log_sync flag. > > Fixing occasional segfaults upon exit in some tests > > Fixed occasional garbled version string. > > Added system integration test which loads 5G of data, restarts the > > cluster and does a scan to verify the load. Fixed infinite in-memory > > AccessGroup compaction loop > > Lazy load bloom filters > > Added rsclient 10x split/merge loop test > > Applied changes to build with GCC 4.4.0 > > Fixed mutator log sync regression test and cleaning up dumplog. > > Added Hypertable.HqlInterpreter.Mutator.NoLogSync property > > Added command line switch --no-log-sync to hypertable cli for setting > > the aforementioned property Fixed a couple of rare NullPointerExceptions > in > > HdfsBroker > > Added flags argument to ThriftClient open_mutator > > Modified load generator tool to support no log sync and flush interval > > flags. Fixed bug in commit log garbage collection that was caused some > > cells to be dropped. Print out bad line number on LOAD DATA INFILE > > exception Fixed ThriftClient-perl regression test > > Send KEYS_ONLY clause over to RangeServer for processing server-side > > Fixed test for Python Thrift module > > Changed RangeServer.MemoryLimit.Percentage default from 80 to 70 > > Added size of block cache to memory tracker > > Got rid of hardcoded 60 second DfsBroker.Timeout > > > > > > > > > > > > > > > > ************************************************** > Assoc. Prof. Milko Todorov Marinov, PhD > Department of Computer Systems & Technologies > University of Rousse > 8 Studentska Str, 7017 Rousse, BULGARIA > Tel. (+359 82) 888 356, Fax. (+359 82) 845 708 > Cell phone: (+359 88) 999 6793 > E-mail: [email protected] > ************************************************* > > > > > --~--~---------~--~----~------------~-------~--~----~ You received this message because you are subscribed to the Google Groups "Hypertable Development" group. To post to this group, send email to [email protected] To unsubscribe from this group, send email to [email protected] For more options, visit this group at http://groups.google.com/group/hypertable-dev?hl=en -~----------~----~----~----~------~----~------~--~---
