[
https://issues.apache.org/jira/browse/HBASE-7667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13885673#comment-13885673
]
Sergey Shelukhin commented on HBASE-7667:
-
the docs were committed to the book as
[
https://issues.apache.org/jira/browse/HBASE-7667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13883479#comment-13883479
]
Andrew Purtell commented on HBASE-7667:
---
Any doc updates available? 0.98.0RC1 is
[
https://issues.apache.org/jira/browse/HBASE-7667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13843340#comment-13843340
]
Sergey Shelukhin commented on HBASE-7667:
-
With 98 coming so soon, probably
[
https://issues.apache.org/jira/browse/HBASE-7667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13843355#comment-13843355
]
stack commented on HBASE-7667:
--
Yeah. Too late for 0.96. We are trying to get back on to a
[
https://issues.apache.org/jira/browse/HBASE-7667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13842713#comment-13842713
]
Otis Gospodnetic commented on HBASE-7667:
-
Btw. is this going to get into any
[
https://issues.apache.org/jira/browse/HBASE-7667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13808236#comment-13808236
]
stack commented on HBASE-7667:
--
bq. Yes, unless there's non-uniform key access there will be
[
https://issues.apache.org/jira/browse/HBASE-7667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13808248#comment-13808248
]
stack commented on HBASE-7667:
--
If I read the user doc., it says:
This improves read
[
https://issues.apache.org/jira/browse/HBASE-7667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13808404#comment-13808404
]
Sergey Shelukhin commented on HBASE-7667:
-
[~stack] I understand your concern. I
[
https://issues.apache.org/jira/browse/HBASE-7667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13806903#comment-13806903
]
Sergey Shelukhin commented on HBASE-7667:
-
I have run the maven tests and rebased
[
https://issues.apache.org/jira/browse/HBASE-7667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13807290#comment-13807290
]
stack commented on HBASE-7667:
--
So, stripe compactions does more i/o unless it is the time
[
https://issues.apache.org/jira/browse/HBASE-7667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13807397#comment-13807397
]
Sergey Shelukhin commented on HBASE-7667:
-
Yes, unless there's non-uniform key
[
https://issues.apache.org/jira/browse/HBASE-7667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13690757#comment-13690757
]
Sergey Shelukhin commented on HBASE-7667:
-
Size-based scheme works by splitting
[
https://issues.apache.org/jira/browse/HBASE-7667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13690761#comment-13690761
]
Sergey Shelukhin commented on HBASE-7667:
-
So it's easier to configure (just say I
[
https://issues.apache.org/jira/browse/HBASE-7667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13688945#comment-13688945
]
stack commented on HBASE-7667:
--
Rereading the design doc and how-to-use. They are very nice.
[
https://issues.apache.org/jira/browse/HBASE-7667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13689954#comment-13689954
]
Sergey Shelukhin commented on HBASE-7667:
-
I'm looking at how to merge policies.
[
https://issues.apache.org/jira/browse/HBASE-7667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13690015#comment-13690015
]
stack commented on HBASE-7667:
--
bq. splitting uses more I/O than not splitting
Sorry. You
[
https://issues.apache.org/jira/browse/HBASE-7667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13654572#comment-13654572
]
Jonathan Hsieh commented on HBASE-7667:
---
[~sershe], thanks for the graph. Seeing
[
https://issues.apache.org/jira/browse/HBASE-7667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13651556#comment-13651556
]
Elliott Clark commented on HBASE-7667:
--
Thanks for the doc. Reading this tonight.
[
https://issues.apache.org/jira/browse/HBASE-7667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13649312#comment-13649312
]
Raymond commented on HBASE-7667:
great, more region lead to better load balance and good
[
https://issues.apache.org/jira/browse/HBASE-7667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13618967#comment-13618967
]
Sergey Shelukhin commented on HBASE-7667:
-
Currently only one compaction per store
[
https://issues.apache.org/jira/browse/HBASE-7667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13619380#comment-13619380
]
Sergey Shelukhin commented on HBASE-7667:
-
Actually, judging by logs what can be
[
https://issues.apache.org/jira/browse/HBASE-7667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13618189#comment-13618189
]
Matt Corgan commented on HBASE-7667:
Sergey - i'm curious how are compactions of the
[
https://issues.apache.org/jira/browse/HBASE-7667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13617951#comment-13617951
]
Sergey Shelukhin commented on HBASE-7667:
-
I did a c1.xlarge test (with default,
[
https://issues.apache.org/jira/browse/HBASE-7667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13617956#comment-13617956
]
Ted Yu commented on HBASE-7667:
---
bq. 5) Get rid of L0?
Can we do this frist ?
[
https://issues.apache.org/jira/browse/HBASE-7667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13616654#comment-13616654
]
Sergey Shelukhin commented on HBASE-7667:
-
I will update the doc, although in this
[
https://issues.apache.org/jira/browse/HBASE-7667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13616976#comment-13616976
]
Sergey Shelukhin commented on HBASE-7667:
-
Btw, the 3 next child JIRAs are rady
[
https://issues.apache.org/jira/browse/HBASE-7667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13614805#comment-13614805
]
Andrew Purtell commented on HBASE-7667:
---
I'd recommend retesting with c1.xlarge
[
https://issues.apache.org/jira/browse/HBASE-7667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13612653#comment-13612653
]
M. C. Srivas commented on HBASE-7667:
-
@mcorgan and @stack:
The total i/o in terms of
[
https://issues.apache.org/jira/browse/HBASE-7667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13612808#comment-13612808
]
stack commented on HBASE-7667:
--
[~srivas]
bq. But the disk iops are much, much worse.
See
[
https://issues.apache.org/jira/browse/HBASE-7667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13612847#comment-13612847
]
Matt Corgan commented on HBASE-7667:
{quote}a RS will write 100x more files into HDFS
[
https://issues.apache.org/jira/browse/HBASE-7667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13612956#comment-13612956
]
Sergey Shelukhin commented on HBASE-7667:
-
bq. The first half of the sentence
[
https://issues.apache.org/jira/browse/HBASE-7667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13613004#comment-13613004
]
stack commented on HBASE-7667:
--
HBASE-7845 optimize hfile index key is the key/boundary
[
https://issues.apache.org/jira/browse/HBASE-7667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13613454#comment-13613454
]
Lars Hofhansl commented on HBASE-7667:
--
bq. An interesting comment by LarsH recently
[
https://issues.apache.org/jira/browse/HBASE-7667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13611728#comment-13611728
]
M. C. Srivas commented on HBASE-7667:
-
There is of course one major caveat with this
[
https://issues.apache.org/jira/browse/HBASE-7667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13611820#comment-13611820
]
Sergey Shelukhin commented on HBASE-7667:
-
There are two approaches discussed, one
[
https://issues.apache.org/jira/browse/HBASE-7667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13611843#comment-13611843
]
stack commented on HBASE-7667:
--
bq. There is of course one major caveat with this approach.
[
https://issues.apache.org/jira/browse/HBASE-7667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13611844#comment-13611844
]
stack commented on HBASE-7667:
--
{code}
On attached doc, it is lovely.
Missing author, date,
[
https://issues.apache.org/jira/browse/HBASE-7667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13611909#comment-13611909
]
Matt Corgan commented on HBASE-7667:
{quote}If data insertion is uniformly spread (ie,
[
https://issues.apache.org/jira/browse/HBASE-7667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13611324#comment-13611324
]
Sergey Shelukhin commented on HBASE-7667:
-
Hmm... I am starting to think that we
[
https://issues.apache.org/jira/browse/HBASE-7667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13609873#comment-13609873
]
Ted Yu commented on HBASE-7667:
---
Nice document, Sergey.
bq. This can obviously be improved
[
https://issues.apache.org/jira/browse/HBASE-7667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13578445#comment-13578445
]
Jimmy Xiang commented on HBASE-7667:
Getting rid of L0 means memstore flushing will
[
https://issues.apache.org/jira/browse/HBASE-7667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13578513#comment-13578513
]
stack commented on HBASE-7667:
--
bq. Getting rid of L0 means memstore flushing will take
[
https://issues.apache.org/jira/browse/HBASE-7667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13578037#comment-13578037
]
Sergey Shelukhin commented on HBASE-7667:
-
If the user copies the file, few things
[
https://issues.apache.org/jira/browse/HBASE-7667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13578087#comment-13578087
]
Matt Corgan commented on HBASE-7667:
Sergey - I'm curious what's the reasoning behind
[
https://issues.apache.org/jira/browse/HBASE-7667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13578090#comment-13578090
]
Sergey Shelukhin commented on HBASE-7667:
-
My reasoning was - too many really tiny
[
https://issues.apache.org/jira/browse/HBASE-7667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13578099#comment-13578099
]
stack commented on HBASE-7667:
--
Would it simplify the filesystem implementation if you did
[
https://issues.apache.org/jira/browse/HBASE-7667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13578105#comment-13578105
]
Matt Corgan commented on HBASE-7667:
Gotcha. Agree about limiting scope. If the
[
https://issues.apache.org/jira/browse/HBASE-7667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13578108#comment-13578108
]
Sergey Shelukhin commented on HBASE-7667:
-
L0 actually should be relatively easy
[
https://issues.apache.org/jira/browse/HBASE-7667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13578192#comment-13578192
]
stack commented on HBASE-7667:
--
I'd like to get rid of L0 so can do splitting w/o resorting
[
https://issues.apache.org/jira/browse/HBASE-7667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13576798#comment-13576798
]
Sergey Shelukhin commented on HBASE-7667:
-
The stripe boundaries are supposed to
[
https://issues.apache.org/jira/browse/HBASE-7667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13576826#comment-13576826
]
stack commented on HBASE-7667:
--
bq. Currently Sergey puts the boundaries in metadata of store
[
https://issues.apache.org/jira/browse/HBASE-7667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13576829#comment-13576829
]
Sergey Shelukhin commented on HBASE-7667:
-
In the N stripes scheme, boundaries
[
https://issues.apache.org/jira/browse/HBASE-7667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13576836#comment-13576836
]
Ted Yu commented on HBASE-7667:
---
bq. It is the bounds of the keys the file contains? (Is
[
https://issues.apache.org/jira/browse/HBASE-7667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13576845#comment-13576845
]
stack commented on HBASE-7667:
--
[~ted_yu] You raise a 'concern' that is nebulous to shoot
[
https://issues.apache.org/jira/browse/HBASE-7667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13576881#comment-13576881
]
Sergey Shelukhin commented on HBASE-7667:
-
Discussed with Ted, what was meant is
[
https://issues.apache.org/jira/browse/HBASE-7667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13577311#comment-13577311
]
Ted Yu commented on HBASE-7667:
---
Here is my understanding of the difference between existing
[
https://issues.apache.org/jira/browse/HBASE-7667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13575883#comment-13575883
]
Nicolas Spiegelberg commented on HBASE-7667:
Some thoughts I had about this:
[
https://issues.apache.org/jira/browse/HBASE-7667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13575936#comment-13575936
]
Sergey Shelukhin commented on HBASE-7667:
-
bq. It may be a follow on to this jira,
[
https://issues.apache.org/jira/browse/HBASE-7667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13575951#comment-13575951
]
stack commented on HBASE-7667:
--
In a striped region, if stripes = 2 and the key distribution
[
https://issues.apache.org/jira/browse/HBASE-7667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13575995#comment-13575995
]
Matt Corgan commented on HBASE-7667:
{quote}Open Times : I think this will be an
[
https://issues.apache.org/jira/browse/HBASE-7667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13576023#comment-13576023
]
Sergey Shelukhin commented on HBASE-7667:
-
bq. a split could just rename the
[
https://issues.apache.org/jira/browse/HBASE-7667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13576044#comment-13576044
]
Jimmy Xiang commented on HBASE-7667:
bq. Wouldn't that be enough to process delete
[
https://issues.apache.org/jira/browse/HBASE-7667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13576296#comment-13576296
]
Ted Yu commented on HBASE-7667:
---
I think we need to find the best way of storing stripe
[
https://issues.apache.org/jira/browse/HBASE-7667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13575449#comment-13575449
]
Ted Yu commented on HBASE-7667:
---
bq. Sub-region is not a good name either though.
Other
[
https://issues.apache.org/jira/browse/HBASE-7667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13575578#comment-13575578
]
Jimmy Xiang commented on HBASE-7667:
Stripes don't have overlapping keyrange with
[
https://issues.apache.org/jira/browse/HBASE-7667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13575583#comment-13575583
]
Lars Hofhansl commented on HBASE-7667:
--
I see, then a major compaction needs to at
[
https://issues.apache.org/jira/browse/HBASE-7667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13575589#comment-13575589
]
Jimmy Xiang commented on HBASE-7667:
That's right. To major compact a stripe, all L0
[
https://issues.apache.org/jira/browse/HBASE-7667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13575206#comment-13575206
]
Jimmy Xiang commented on HBASE-7667:
bq. Overall, I think the sub-region approach cuts
[
https://issues.apache.org/jira/browse/HBASE-7667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13575297#comment-13575297
]
stack commented on HBASE-7667:
--
[~mcorgan] Nice write up. So you are in favor of Sergey's
[
https://issues.apache.org/jira/browse/HBASE-7667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13575309#comment-13575309
]
Matt Corgan commented on HBASE-7667:
{quote}So you are in favor of Sergey's
[
https://issues.apache.org/jira/browse/HBASE-7667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13575312#comment-13575312
]
Lars Hofhansl commented on HBASE-7667:
--
Stripes can have overlapping keyrange with
[
https://issues.apache.org/jira/browse/HBASE-7667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13575012#comment-13575012
]
Jimmy Xiang commented on HBASE-7667:
A stripe is like a sub-region. That's a good
[
https://issues.apache.org/jira/browse/HBASE-7667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13575024#comment-13575024
]
Matt Corgan commented on HBASE-7667:
Right now there is a tug-of-war between region
[
https://issues.apache.org/jira/browse/HBASE-7667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13562257#comment-13562257
]
Sergey Shelukhin commented on HBASE-7667:
-
What do you guys think about the
[
https://issues.apache.org/jira/browse/HBASE-7667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13562268#comment-13562268
]
Matt Corgan commented on HBASE-7667:
So a stripe is like a sub-region? In terms of
[
https://issues.apache.org/jira/browse/HBASE-7667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13562273#comment-13562273
]
chunhui shen commented on HBASE-7667:
-
Interesting ideas.
With stripe compaction, we
[
https://issues.apache.org/jira/browse/HBASE-7667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13562278#comment-13562278
]
Sergey Shelukhin commented on HBASE-7667:
-
Hmm, my thinking would be that number
[
https://issues.apache.org/jira/browse/HBASE-7667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13562286#comment-13562286
]
Sergey Shelukhin commented on HBASE-7667:
-
If there are no general objections I
[
https://issues.apache.org/jira/browse/HBASE-7667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13562290#comment-13562290
]
Matt Corgan commented on HBASE-7667:
I've been brainstorming something similar for
79 matches
Mail list logo