[
https://issues.apache.org/jira/browse/HBASE-19285?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16257723#comment-16257723
]
Josh Elser commented on HBASE-19285:
------------------------------------
Well, color me surprised. Initial tests show no impact. This makes me think I
did something wrong.
{noformat}
$ bin/hbase pe --latency --nomapred --presplit=1000 --valueSize=1000
--rows=100000 sequentialWrite 30
{noformat}
{noformat}
### With table metrics
2017-11-17 22:36:53,939 INFO [main] hbase.PerformanceEvaluation:
[SequentialWriteTest] Summary of timings (ms): [59063, 59046, 60074, 59116,
59436, 59829, 58789, 59049, 60074, 59087, 58798, 59934, 59328, 59325, 59790,
59317, 59961, 60116, 58530, 59394, 59515, 58430, 58510, 59290, 59376, 58812,
59762, 58578, 59929, 59388]
2017-11-17 22:36:53,940 INFO [main] hbase.PerformanceEvaluation:
[SequentialWriteTest] Min: 58430ms Max: 60116ms Avg: 59321ms
2017-11-17 22:40:24,731 INFO [main] hbase.PerformanceEvaluation:
[SequentialWriteTest] Summary of timings (ms): [56313, 56755, 56521, 52565,
54266, 53052, 56263, 54206, 56329, 51822, 52421, 54228, 53171, 56686, 56707,
53018, 50557, 56529, 56326, 56704, 56519, 54258, 56503, 54292, 56338, 53167,
56508, 53116, 52534, 56193]
2017-11-17 22:40:24,733 INFO [main] hbase.PerformanceEvaluation:
[SequentialWriteTest] Min: 50557ms Max: 56755ms Avg: 54795ms
2017-11-17 22:44:45,189 INFO [main] hbase.PerformanceEvaluation:
[SequentialWriteTest] Summary of timings (ms): [64965, 63050, 63186, 63057,
63068, 63450, 63929, 63121, 63272, 62800, 64640, 64476, 63888, 62953, 64284,
62958, 64466, 64607, 64266, 63359, 64628, 64374, 64948, 63883, 64322, 63837,
63264, 64348, 64670, 63408]
2017-11-17 22:44:45,190 INFO [main] hbase.PerformanceEvaluation:
[SequentialWriteTest] Min: 62800ms Max: 64965ms Avg: 63849ms
### Without table metrics (stock)
2017-11-17 22:52:01,694 INFO [main] hbase.PerformanceEvaluation:
[SequentialWriteTest] Summary of timings (ms): [68170, 68175, 61443, 69265,
69226, 69447, 69371, 60942, 69250, 69088, 65739, 65723, 65756, 65764, 63096,
68151, 65782, 63133, 68174, 68193, 63076, 61539, 65809, 68167, 68984, 69373,
67716, 68154, 65756, 69024]
2017-11-17 22:52:01,694 INFO [main] hbase.PerformanceEvaluation:
[SequentialWriteTest] Min: 60942ms Max: 69447ms Avg: 66716ms
2017-11-17 22:57:35,774 INFO [main] hbase.PerformanceEvaluation:
[SequentialWriteTest] Summary of timings (ms): [62112, 61847, 62267, 61856,
62088, 61863, 59649, 61915, 62237, 60353, 59469, 60340, 59375, 60396, 59450,
59625, 62170, 60338, 59000, 61362, 60308, 61921, 61300, 61834, 62142, 59622,
62203, 61935, 59639, 59467]
2017-11-17 22:57:35,775 INFO [main] hbase.PerformanceEvaluation:
[SequentialWriteTest] Min: 59000ms Max: 62267ms Avg: 60936ms
2017-11-17 23:00:28,712 INFO [main] hbase.PerformanceEvaluation:
[SequentialWriteTest] Summary of timings (ms): [58838, 64460, 63499, 65246,
59610, 60985, 63585, 60948, 65323, 58409, 58478, 63565, 63601, 61111, 64464,
65149, 60909, 59680, 65169, 61050, 65371, 65197, 59647, 60922, 59605, 65232,
63591, 65147, 65307, 64475]
2017-11-17 23:00:28,713 INFO [main] hbase.PerformanceEvaluation:
[SequentialWriteTest] Min: 58409ms Max: 65371ms Avg: 62619ms
{noformat}
Single RS, 8G heap a couple of memstore flush config tweaks to try to smooth
out the 1K regions on a single RS. My only guess is that I'm still getting
blocked on flushes which is dominating execution time. Let me poke with JFR or
YourKit.
> Add per-table latency histograms
> --------------------------------
>
> Key: HBASE-19285
> URL: https://issues.apache.org/jira/browse/HBASE-19285
> Project: HBase
> Issue Type: Bug
> Components: metrics
> Reporter: Clay B.
> Assignee: Josh Elser
> Priority: Critical
> Fix For: 2.0.0, 1.4.0, 1.3.3
>
>
> HBASE-17017 removed the per-region latency histograms (e.g. Get, Put, Scan at
> p75, p85, etc)
> HBASE-15518 added some per-table metrics, but not the latency histograms.
> Given the previous conversations, it seems like it these per-table
> aggregations weren't intentionally omitted, just never re-implemented after
> the per-region removal. They're some really nice out-of-the-box metrics we
> can provide to our users/admins as long as it's not detrimental.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
