Displaying #cassandra-dev/2017-05-31.log:

Wed May 31 00:07:26 2017  kohlisankalp:Joined the channel
Wed May 31 00:39:07 2017  RussSpitzer:Joined the channel
Wed May 31 00:42:25 2017  kohlisankalp:Joined the channel
Wed May 31 00:56:43 2017  aboudreault:Joined the channel
Wed May 31 02:00:52 2017  dikang:Joined the channel
Wed May 31 02:31:44 2017  kohlisankalp:Joined the channel
Wed May 31 02:51:08 2017  dikang:Joined the channel
Wed May 31 03:33:49 2017  krut:i thought LCS was meant to do STCS on unrepaired sstables, but i can't figure out where that happens in the code. it appears that the repaired and unrepaired pools are always LCS, and i can't find anything that does anything sneaky like force all sstables to be in L0 for unrepaired. have i been lied to my whole life or am i just looking in the wrong places?
Wed May 31 03:34:20 2017  zanson:it did that briefly, but it changed to LCS everything
Wed May 31 03:34:43 2017  zanson:because STCS unrepaired could get you in trouble if you weren't running repair often enough
Wed May 31 03:35:15 2017  krut:that makes sense. when did it change?
Wed May 31 03:35:37 2017  zanson:forget, you would have to jira dive or git blame around to figure it out
Wed May 31 03:35:50 2017  kohlisankalp:Joined the channel
Wed May 31 03:36:24 2017  krut:eh i'm not that interested. would have probably been an earlier version of 2.1 i'm guessing
Wed May 31 03:37:16 2017  dikang:Joined the channel
Wed May 31 05:25:50 2017  kvaster:Joined the channel
Wed May 31 05:37:13 2017  cosql:Joined the channel
Wed May 31 05:40:17 2017  dikang:Joined the channel
Wed May 31 06:08:30 2017  marcuse:krut: https://issues.apache.org/jira/browse/CASSANDRA-8004
Wed May 31 06:59:16 2017  spodkowinski:Joined the channel
Wed May 31 07:29:22 2017  kohlisankalp:Joined the channel
Wed May 31 07:32:17 2017  kohlisankalp:Joined the channel
Wed May 31 08:19:34 2017  makrusak:Joined the channel
Wed May 31 08:27:27 2017  cosql:Joined the channel
Wed May 31 08:52:39 2017  clohfink_:Joined the channel
Wed May 31 09:06:23 2017  RussSpitzer:Joined the channel
Wed May 31 09:08:27 2017  echoreply:Joined the channel
Wed May 31 09:32:48 2017  kvaster:Joined the channel
Wed May 31 09:41:13 2017  kvaster_:Joined the channel
Wed May 31 09:53:58 2017  jmckenzie_:Joined the channel
Wed May 31 10:02:39 2017  kvaster_:Joined the channel
Wed May 31 10:50:04 2017  ra21vi:Joined the channel
Wed May 31 12:50:49 2017  romainh:Joined the channel
Wed May 31 13:19:05 2017  jasobrown:what's the magic foo for eagerly cleaning up a direct bytebuffer?
Wed May 31 13:19:26 2017  jasobrown:i thought we had a helper function to execute that somewhere ...
Wed May 31 13:21:30 2017  jasobrown:ahhh, i think this is the magic: ((DirectBuffer) buf).cleaner().clean()
Wed May 31 13:29:31 2017  iamaleksey:we do
Wed May 31 13:29:38 2017  iamaleksey:in bbutils
Wed May 31 13:30:11 2017  jasobrown:lol - it's actually in FileUtils
Wed May 31 13:30:19 2017  jasobrown:public static void clean(ByteBuffer buffer)
Wed May 31 13:30:26 2017  jasobrown:that's why i couldn't find it
Wed May 31 13:30:30 2017  iamaleksey:oh; or there
Wed May 31 13:53:28 2017  jasobrown:as a general rule, i should cleanup and direct bytebuffers i allocate, yes?
Wed May 31 13:53:40 2017  jasobrown:to release the native memory faster
Wed May 31 14:19:48 2017  jmckenzie:Joined the channel
Wed May 31 14:32:58 2017  tjake:yes, though it has a cost
Wed May 31 14:33:06 2017  tjake:ideally you should pool them
Wed May 31 14:46:35 2017  dikang:Joined the channel
Wed May 31 14:51:06 2017  iamaleksey:contends on at least two locks, right? to update total direct bb memory allocated and for the cleaners queue?
Wed May 31 14:51:30 2017  dikang:Joined the channel
Wed May 31 15:15:42 2017  tjake:Netty has a way to allocate them without a cleaner
Wed May 31 15:15:47 2017  tjake:might want to use that
Wed May 31 15:16:32 2017  tjake:PlatformDependent.allocateDirectNoCleaner
Wed May 31 15:20:38 2017  dikang_:Joined the channel
Wed May 31 15:22:24 2017  ra21vi:I read secondary index isnt good for query that spans multi node. Does SASI index overcome this?
Wed May 31 15:22:41 2017  ra21vi:whats the real benefit of SASI index which is released currently
Wed May 31 15:37:44 2017  clohfink:Joined the channel
Wed May 31 15:50:35 2017  romainh1:Joined the channel
Wed May 31 15:51:35 2017  kohlisankalp:Joined the channel
Wed May 31 17:05:25 2017  kohlisankalp:Joined the channel
Wed May 31 17:10:25 2017  jasobrown:iamaleksey: what contends on two locks? our BufferPool impl?
Wed May 31 17:10:40 2017  jasobrown:tjake: thanks for the netty tip ;) didn't know that trick
Wed May 31 17:11:52 2017  kohlisankalp:Joined the channel
Wed May 31 17:18:27 2017  dikang:Joined the channel
Wed May 31 17:18:54 2017  jasobrown:iamaleksey & tjake: just walked through the BB.allocateDirect path, and yeah I see the locks and a crap ton of volatile writes
Wed May 31 17:19:07 2017  jasobrown:pooling/the netty things makes sense
Wed May 31 17:19:14 2017  jasobrown:netty thing*
Wed May 31 17:20:03 2017  jasobrown:thx for the pointers!
Wed May 31 17:21:31 2017  tjake:or use the BufferPool
Wed May 31 17:21:57 2017  dikang:Joined the channel
Wed May 31 17:22:59 2017  jasobrown:tjake: yup, switching to use that now
Wed May 31 17:41:09 2017  cosql:Joined the channel
Wed May 31 17:55:17 2017  dikang_:Joined the channel
Wed May 31 19:02:20 2017  mebigfatguy:Joined the channel
Wed May 31 20:03:53 2017  kohlisan_:Joined the channel
Wed May 31 20:16:34 2017  wadey:Joined the channel
Wed May 31 20:28:13 2017  mebigfatguy:Joined the channel
Wed May 31 20:30:24 2017  kohlisankalp:Joined the channel
Wed May 31 20:51:13 2017  kohlisankalp:Joined the channel
Wed May 31 20:51:40 2017  dikang:Joined the channel
Wed May 31 22:11:20 2017  dikang_:Joined the channel
Wed May 31 22:28:59 2017  clohfink:Joined the channel
Wed May 31 22:30:14 2017  clohfink:Joined the channel
Wed May 31 22:55:49 2017  clohfink:Joined the channel
Wed May 31 22:57:58 2017  kohlisankalp:Joined the channel
Wed May 31 22:59:49 2017  clohfink:Joined the channel
Wed May 31 23:05:57 2017  clohfink:Joined the channel
Wed May 31 23:42:53 2017  bbromhead:Joined the channel
Wed May 31 23:44:05 2017  bbromhead:any folks going to be in SF for spark summit?
Wed May 31 23:44:45 2017  driftx:I'm not sure if RussSpitzer's auto alert for 'spark' works in here
Wed May 31 23:44:54 2017  RussSpitzer:It does :)
Wed May 31 23:44:57 2017  RussSpitzer:and I will be
Wed May 31 23:44:57 2017  driftx:lol
Wed May 31 23:45:00 2017  mebigfatguy:heh
Wed May 31 23:45:15 2017  RussSpitzer:my talk is on Wed at 2 I think
Wed May 31 23:45:19 2017  RussSpitzer:let me grab a link
Wed May 31 23:45:45 2017  RussSpitzer:https://spark-summit.org/2017/events/cassandra-and-sparksql-you-dont-need-functional-programming-for-fun/
Wed May 31 23:46:29 2017  bbromhead:awesome, I will check it out
Wed May 31 23:46:46 2017  bbromhead:also open invitation for any cassandra folk to grab a beer!

Comments