Displaying #cassandra-dev/2017-01-13.log:

Fri Jan 13 00:01:58 2017  dikang:Joined the channel
Fri Jan 13 00:22:49 2017  kohlisankalp:Joined the channel
Fri Jan 13 00:32:47 2017  mstepura:Joined the channel
Fri Jan 13 00:43:03 2017  dikang:Joined the channel
Fri Jan 13 00:45:28 2017  kohlisankalp:Joined the channel
Fri Jan 13 01:03:21 2017  dikang:Joined the channel
Fri Jan 13 01:12:41 2017  anthony:Joined the channel
Fri Jan 13 01:16:07 2017  kohlisankalp:Joined the channel
Fri Jan 13 01:32:23 2017  kohlisankalp:Joined the channel
Fri Jan 13 01:35:29 2017  rustyrazorblade:sad face
Fri Jan 13 02:06:51 2017  jmckenzie_:Joined the channel
Fri Jan 13 02:14:23 2017  anthony:Joined the channel
Fri Jan 13 02:17:19 2017  anthony:Joined the channel
Fri Jan 13 02:29:54 2017  RussSpitzer:Joined the channel
Fri Jan 13 02:36:29 2017  rustyrazorblade:still down
Fri Jan 13 02:49:28 2017  mstepura:Joined the channel
Fri Jan 13 02:58:41 2017  mstepura:Joined the channel
Fri Jan 13 03:14:51 2017  jmckenzie:Joined the channel
Fri Jan 13 03:19:29 2017  mstepura:Joined the channel
Fri Jan 13 03:49:40 2017  kohlisankalp:Joined the channel
Fri Jan 13 04:03:44 2017  kohlisankalp:Joined the channel
Fri Jan 13 04:07:01 2017  RussSpitzer:Joined the channel
Fri Jan 13 04:24:41 2017  anthony:Joined the channel
Fri Jan 13 04:31:52 2017  clohfink:Joined the channel
Fri Jan 13 04:35:03 2017  anthony_:Joined the channel
Fri Jan 13 04:43:32 2017  mstepura:Joined the channel
Fri Jan 13 04:44:27 2017  mstepura:Joined the channel
Fri Jan 13 04:45:11 2017  mstepura:Joined the channel
Fri Jan 13 04:45:59 2017  mstepura:Joined the channel
Fri Jan 13 04:46:49 2017  mstepura:Joined the channel
Fri Jan 13 04:47:39 2017  mstepura:Joined the channel
Fri Jan 13 04:48:25 2017  mstepura:Joined the channel
Fri Jan 13 04:49:15 2017  mstepura:Joined the channel
Fri Jan 13 04:49:55 2017  mstepura:Joined the channel
Fri Jan 13 05:00:20 2017  joaquin:Joined the channel
Fri Jan 13 05:04:36 2017  joaquin:Joined the channel
Fri Jan 13 05:33:53 2017  joaquin:Joined the channel
Fri Jan 13 05:39:40 2017  dikang:Joined the channel
Fri Jan 13 06:01:36 2017  kvaster:Joined the channel
Fri Jan 13 06:18:32 2017  maedhroz:Joined the channel
Fri Jan 13 06:24:35 2017  jfarrell:Joined the channel
Fri Jan 13 06:29:18 2017  joaquin:Joined the channel
Fri Jan 13 06:54:12 2017  mstepura:Joined the channel
Fri Jan 13 07:31:06 2017  rsmags_:Joined the channel
Fri Jan 13 07:31:12 2017  mck_sw:Joined the channel
Fri Jan 13 07:31:19 2017  jbellis_:Joined the channel
Fri Jan 13 07:32:27 2017  jrwest_:Joined the channel
Fri Jan 13 07:35:34 2017  ostefano:Joined the channel
Fri Jan 13 07:36:50 2017  yukim:Joined the channel
Fri Jan 13 07:45:46 2017  cosql:Joined the channel
Fri Jan 13 08:08:49 2017  spodkowinski:Joined the channel
Fri Jan 13 08:18:58 2017  kvaster:Joined the channel
Fri Jan 13 08:38:34 2017  cosql:Joined the channel
Fri Jan 13 09:17:07 2017  anthon___:Joined the channel
Fri Jan 13 09:49:24 2017  mebigfatguy:Joined the channel
Fri Jan 13 13:37:52 2017  stephenc:Joined the channel
Fri Jan 13 13:46:54 2017  jmckenzie_:Joined the channel
Fri Jan 13 13:59:13 2017  anthony__:Joined the channel
Fri Jan 13 14:10:54 2017  adamholmberg:Joined the channel
Fri Jan 13 14:48:47 2017  zaller:Joined the channel
Fri Jan 13 14:49:24 2017  aweisberg:Still down, and the status page still says everything is working?
Fri Jan 13 14:49:36 2017  iamaleksey:works for me, just is slow
Fri Jan 13 14:53:37 2017  aweisberg:Huh, if I follow the link from issues.apache.org it is down, but if I refresh it is up
Fri Jan 13 15:44:10 2017  zaller:Joined the channel
Fri Jan 13 15:50:10 2017  joaquin:Joined the channel
Fri Jan 13 16:05:12 2017  adamholmberg:Joined the channel
Fri Jan 13 16:36:03 2017  thobbs:Joined the channel
Fri Jan 13 16:44:27 2017  aweisberg:In this snippet how exact does column k go from null to k1 http://pastebin.com/v8qmzWSA
Fri Jan 13 16:44:27 2017  aweisberg:from the statement 'UPDATE test SET v='foo', version=1 WHERE id=0 AND k='k1' IF version = 0'
Fri Jan 13 16:49:47 2017  mstepura:Joined the channel
Fri Jan 13 16:54:47 2017  zanson:aweisberg: looks like a bug to me
Fri Jan 13 16:54:59 2017  aweisberg:This is from a test and the test expects this result.
Fri Jan 13 16:55:04 2017  zanson:null clustering column isn't being kept as valid
Fri Jan 13 16:55:48 2017  zanson:really? that seems like a bad test. does it do that on 2.1 and 3.x?
Fri Jan 13 16:55:59 2017  aweisberg:https://github.com/riptano/cassandra-dtest/blob/master/upgrade_tests/cql_tests.py#L3788
Fri Jan 13 16:56:41 2017  aweisberg:It's an upgrade test and that particular thing passes on all versions.
Fri Jan 13 16:56:45 2017  zanson:because my read of that is you should have a (0,null) row and a (0,k1) row
Fri Jan 13 16:57:04 2017  zanson:but maybe I am missing something? ¯\_(ツ)_/¯
Fri Jan 13 16:57:22 2017  aweisberg:It's a fun test. Static columns are very interesting.
Fri Jan 13 16:58:27 2017  zanson:I would expect two rows. 0,null,1,null and 0,k1,1,foo to be the rows after those actions
Fri Jan 13 17:01:24 2017  jmckenzie:Joined the channel
Fri Jan 13 17:02:48 2017  maedhroz:Joined the channel
Fri Jan 13 17:38:12 2017  anthony_:Joined the channel
Fri Jan 13 17:40:51 2017  jkni:That's how I'd expect it to work (having worked with static columns)
Fri Jan 13 17:41:03 2017  jkni:whether that's documented in the CQL spec and elsewhere is another question
Fri Jan 13 17:41:23 2017  jkni:in the language of the 3.0+ storage engine, the first insert generated a partition with key 0 and a static_block, but no rows
Fri Jan 13 17:41:49 2017  jkni:the null in the first select doesn't indicate a row with a null clustering value, but rather the existing of the static block
Fri Jan 13 17:42:14 2017  jkni:the update does a conditional on the static block and updates the static block + creates the row with clusting 'k1' and v='foo' in the partition
Fri Jan 13 17:42:27 2017  aweisberg:This is an area of pretty hot discussion between ifesdjeen and pcmanus https://issues.apache.org/jira/browse/CASSANDRA-12060 https://issues.apache.org/jira/browse/CASSANDRA-12694 I'm trying to decipher it now.
Fri Jan 13 17:42:50 2017  jkni:Yeah, whether I can explain the behavior and whether it is the right behavior are two different questions :)
Fri Jan 13 17:43:31 2017  aweisberg:And you are saying it shouldn't create the row with this clustering because of course the clustering is only in the condition?
Fri Jan 13 17:45:36 2017  aboudreault_:Joined the channel
Fri Jan 13 17:45:47 2017  jkni:I might have deeper issues than that
Fri Jan 13 17:46:02 2017  jkni:I'm not sure that the CQL consistently defines what a row is
Fri Jan 13 17:46:57 2017  jkni:I think the CQL spec also states that an INSERT always requires all columns of the primary key, so we might already be under-specified by the time we get to the first insert
Fri Jan 13 17:47:03 2017  aweisberg:You can update the static block and some number of rows in the same update right?
Fri Jan 13 17:47:08 2017  jkni:yes
Fri Jan 13 17:49:19 2017  dikang:Joined the channel
Fri Jan 13 17:49:44 2017  aweisberg:OK, I think fixing this is above my pay grade.
Fri Jan 13 17:50:00 2017  jkni:I think the complicating factor here is that the first insert generates a row in some sense (a SELECT statement returns rows from a table, after all), but not in terms of an actual storage row with a 'null'
Fri Jan 13 17:57:45 2017  kohlisankalp:Joined the channel
Fri Jan 13 18:00:28 2017  kohlisankalp:Joined the channel
Fri Jan 13 18:01:58 2017  dikang:Joined the channel
Fri Jan 13 18:05:13 2017  dikang:Joined the channel
Fri Jan 13 18:17:53 2017  kohlisankalp:Joined the channel
Fri Jan 13 18:24:55 2017  rustyrazorblade:it pains me to see why that makes sense.
Fri Jan 13 18:29:51 2017  kohlisankalp:Joined the channel
Fri Jan 13 18:34:30 2017  kohlisankalp:Joined the channel
Fri Jan 13 18:38:08 2017  JayZhuang:Joined the channel
Fri Jan 13 18:38:28 2017  thobbs:rows exist if 1) there's a row marker (a special cell in 2.x, something slightly different in 3.x), or 2) there's a live cell
Fri Jan 13 18:38:57 2017  thobbs:the only thing that's slightly confusing about that example is the result of the first query with the "null" clustering column
Fri Jan 13 18:39:10 2017  thobbs:there's no row, only a static block
Fri Jan 13 18:39:37 2017  thobbs:there's not a great way to convey that other than having a null clustering column and values
Fri Jan 13 18:39:58 2017  thobbs:in the second query result, the row actually exists (because there's a live cell in the row)
Fri Jan 13 18:59:19 2017  jkni:Agreed. I guess the downside of not returning a row for the first select is that you end up with no wait to tell if you have a static block
Fri Jan 13 18:59:24 2017  jkni:way*
Fri Jan 13 19:00:19 2017  kvaster:Joined the channel
Fri Jan 13 19:03:03 2017  dikang:Joined the channel
Fri Jan 13 19:04:24 2017  JayZhuang1:Joined the channel
Fri Jan 13 19:06:44 2017  CassBotJr:Joined the channel
Fri Jan 13 19:16:21 2017  JayZhuang:Joined the channel
Fri Jan 13 19:33:14 2017  JayZhuang:Joined the channel
Fri Jan 13 19:39:48 2017  anthony_:Joined the channel
Fri Jan 13 19:46:24 2017  jeffj:anyone aware of any junit tests that actually wait for the tested instance to bootstrap before running tests? havent found any with a quick grep
Fri Jan 13 19:46:27 2017  adamholmberg:Joined the channel
Fri Jan 13 20:00:23 2017  mstepura:Joined the channel
Fri Jan 13 20:32:39 2017  adamholmberg:Joined the channel
Fri Jan 13 20:39:21 2017  mstepura:Joined the channel
Fri Jan 13 20:46:48 2017  joaquin:Joined the channel
Fri Jan 13 20:50:23 2017  kohlisankalp:Joined the channel
Fri Jan 13 20:58:36 2017  marcuse:Joined the channel
Fri Jan 13 20:59:35 2017  kohlisankalp:Joined the channel
Fri Jan 13 21:16:05 2017  joaquin:Joined the channel
Fri Jan 13 21:17:17 2017  joaquin:Joined the channel
Fri Jan 13 21:20:23 2017  zanson:bootstrap? I don't think any junit tests start multiple nodes such that somnething could bootstrap
Fri Jan 13 21:20:33 2017  zanson:those would all be dtests
Fri Jan 13 21:21:55 2017  joaquin:Joined the channel
Fri Jan 13 21:25:04 2017  zanson:jeffj: ^
Fri Jan 13 21:43:58 2017  anthony_:Joined the channel
Fri Jan 13 21:46:54 2017  joaquin:Joined the channel
Fri Jan 13 21:49:53 2017  dikang:Joined the channel
Fri Jan 13 22:02:25 2017  JayZhuang:Joined the channel
Fri Jan 13 22:05:35 2017  dikang:Joined the channel
Fri Jan 13 22:09:28 2017  maedhroz:Joined the channel
Fri Jan 13 22:15:41 2017  anthony_:Joined the channel
Fri Jan 13 22:27:28 2017  jeffj:zanson: you're right, of course - it starts a single instance with auto_bootstrap=false, but the node itself doesnt know that it's "bootstrap complete" immediately
Fri Jan 13 22:40:08 2017  JayZhuang:Joined the channel
Fri Jan 13 23:09:44 2017  kohlisankalp:Joined the channel
Fri Jan 13 23:22:02 2017  JayZhuang:Joined the channel
Fri Jan 13 23:27:56 2017  dikang:Joined the channel
Fri Jan 13 23:28:51 2017  jeffj:zanson: fwiw, CQLTester.requireNetwork() is the magic i was hoping existed.
Fri Jan 13 23:29:04 2017  jeffj:or at least, accomplishes what i wanted to accomplish.
Fri Jan 13 23:29:19 2017  zanson::) cool.
Fri Jan 13 23:29:40 2017  zanson:gotcha, so you didn't actually want to bootstrap something, you just wanted to know when all the startup stuff was done
Fri Jan 13 23:30:31 2017  jeffj:right
Fri Jan 13 23:32:31 2017  zanson:well, actually, to run the startup stuff
Fri Jan 13 23:32:38 2017  zanson::)
Fri Jan 13 23:33:02 2017  zanson:startup finishing stuff.
Fri Jan 13 23:33:26 2017  zanson:cql tester just doesn't actually every finish startup if the test doesn't need to use the network

Comments