-
Type: Bug
-
Resolution: Done
-
Priority: Major - P3
-
Affects Version/s: None
-
Component/s: None
-
None
-
ALL
-
(copied to CRM)
When MongoDB flushes to disk, the rate of inserts fall considerably with at least a 30% reduction.
All the data is in memory and there are no page faults.
The issues persists despite the o/s being highly modified with recommended MongoDB production settings - noatime, ext4, high ulimit, journal on separate disk,ssd and low readahead.
mongostat snippet below showing the drop in query rates when a flush occurs:
insert query update delete getmore command flushes mapped vsize res non-mapped faults locked db idx miss % qr|qw ar|aw netIn netOut conn set repl time 73 6713 6715 0 4859 6806 0 6.47g 13.6g 4.41g 7.12g 0 test:56.7% 0 1|0 0|1 31m 93m 50 rs1 PRI 15:00:00 54 4392 4381 0 3320 4431 0 6.47g 13.6g 4.42g 7.12g 0 test:58.6% 0 39|1 0|1 20m 60m 50 rs1 PRI 15:00:01 64 6078 6113 0 4569 6183 1 6.47g 13.6g 4.4g 7.12g 0 test:67.4% 0 0|0 0|0 29m 84m 50 rs1 PRI 15:00:02 65 5967 5954 0 4549 6020 0 6.47g 13.6g 4.38g 7.12g 0 test:52.3% 0 0|0 0|0 28m 82m 50 rs1 PRI 15:00:03 58 6192 6177 0 4748 6227 0 6.47g 13.6g 4.4g 7.12g 0 test:53.2% 0 39|0 0|1 29m 85m 50 rs1 PRI 15:00:04 65 6212 6246 0 4704 6321 0 6.47g 13.6g 4.38g 7.12g 0 test:53.9% 0 0|0 0|0 29m 86m 50 rs1 PRI 15:00:05 66 6355 6348 0 4957 6412 0 6.47g 13.6g 4.41g 7.12g 0 test:54.8% 0 1|0 0|1 30m 88m 50 rs1 PRI 15:00:06 61 6504 6487 0 4920 6539 0 6.47g 13.6g 4.39g 7.12g 0 test:55.3% 0 39|0 0|1 30m 90m 50 rs1 PRI 15:00:07 63 6654 6683 0 5006 6756 0 6.47g 13.6g 4.37g 7.12g 0 test:57.6% 0 0|0 0|0 31m 92m 50 rs1 PRI 15:00:08 65 6754 6746 0 4928 6811 0 6.47g 13.6g 4.4g 7.12g 0 test:58.0% 0 0|0 0|1 31m 93m 50 rs1 PRI 15:00:09 insert query update delete getmore command flushes mapped vsize res non-mapped faults locked db idx miss % qr|qw ar|aw netIn netOut conn set repl time 46 6166 6154 0 4524 6187 0 6.47g 13.6g 4.38g 7.12g 0 test:51.5% 0 39|0 0|1 29m 85m 50 rs1 PRI 15:00:10 40 3658 3679 0 2882 3734 1 6.47g 13.6g 4.38g 7.12g 0 test:43.8% 0 0|0 0|1 17m 51m 50 rs1 PRI 15:00:11 53 5821 5821 0 4196 5872 0 6.47g 13.6g 4.39g 7.12g 0 test:50.7% 0 4|0 1|1 27m 80m 50 rs1 PRI 15:00:12 64 6370 6366 0 4694 6432 0 6.47g 13.6g 4.38g 7.12g 0 test:55.0% 0 0|0 0|0 30m 88m 50 rs1 PRI 15:00:13 70 6332 6332 0 4794 6403 0 6.47g 13.6g 4.41g 7.12g 0 test:55.0% 0 0|0 0|1 30m 87m 50 rs1 PRI 15:00:14 66 6219 6209 0 4576 6265 0 6.47g 13.6g 4.39g 7.12g 0 test:53.7% 0 39|0 0|1 29m 86m 50 rs1 PRI 15:00:15 75 6151 6172 0 4823 6257 0 6.47g 13.6g 4.37g 7.12g 0 test:53.6% 0 0|0 0|1 29m 85m 50 rs1 PRI 15:00:16 65 6189 6189 0 4697 6255 0 6.47g 13.6g 4.39g 7.12g 0 test:53.3% 0 0|0 0|1 29m 85m 50 rs1 PRI 15:00:17 68 5998 5982 0 4509 6040 0 6.47g 13.6g 4.38g 7.12g 0 test:52.2% 0 39|0 0|1 28m 83m 50 rs1 PRI 15:00:18 62 5975 6001 0 4395 6074 0 6.47g 13.6g 4.4g 7.12g 0 test:54.9% 0 0|0 0|1 28m 83m 50 rs1 PRI 15:00:19 insert query update delete getmore command flushes mapped vsize res non-mapped faults locked db idx miss % qr|qw ar|aw netIn netOut conn set repl time 60 6213 6208 0 4744 6268 0 6.47g 13.6g 4.39g 7.12g 0 test:54.0% 0 1|0 0|1 29m 86m 50 rs1 PRI 15:00:20 43 3738 3732 0 2931 3776 1 6.47g 13.6g 4.38g 7.12g 0 test:37.9% 0 10|0 0|1 17m 51m 50 rs1 PRI 15:00:21 71 6220 6231 0 4783 6302 0 6.47g 13.6g 4.41g 7.12g 0 test:53.5% 0 1|0 1|0 29m 86m 50 rs1 PRI 15:00:22 70 6073 6073 0 4737 6144 0 6.47g 13.6g 4.39g 7.12g 0 test:59.9% 0 8|0 0|1 28m 84m 50 rs1 PRI 15:00:23 57 6608 6619 0 4928 6675 0 6.47g 13.6g 4.37g 7.12g 0 test:58.2% 0 0|0 0|1 31m 91m 50 rs1 PRI 15:00:24 70 6749 6742 0 5221 6814 0 6.47g 13.6g 4.38g 7.12g 0 test:58.2% 0 1|0 0|1 31m 93m 50 rs1 PRI 15:00:25 59 6423 6428 0 4897 6487 0 6.47g 13.6g 4.41g 7.12g 0 test:55.5% 0 6|1 1|1 30m 89m 50 rs1 PRI 15:00:26 72 6294 6306 0 4729 6379 0 6.47g 13.6g 4.38g 7.12g 0 test:57.5% 0 0|0 0|0 29m 87m 50 rs1 PRI 15:00:27 59 6706 6699 0 5002 6756 0 6.47g 13.6g 4.4g 7.12g 0 test:56.6% 0 0|0 0|0 31m 92m 50 rs1 PRI 15:00:28 61 6678 6682 0 4900 6743 0 6.47g 13.6g 4.39g 7.12g 0 test:57.6% 0 1|0 1|0 31m 92m 50 rs1 PRI 15:00:29
- is duplicated by
-
SERVER-12216 Database freezes during flushing mmaps, flushing takes over half minute
- Closed
-
SERVER-6789 Improve performance of flushing data files when using lots of databases/data files
- Closed
- is related to
-
SERVER-5831 server stops processing requests during periodic flush
- Closed
- related to
-
SERVER-12880 Server pauses on requests every 60 seconds
- Closed