Summary:
To make sure `TimedPut` are placed on proper tier before and when it becomes eligible for cold tier
1) flush and compaction need to keep relevant seqno to time mapping for not just the sequence number contained in internal keys, but also preferred sequence number for `TimedPut` entries.
This PR also fix some bugs in for handling `TimedPut` during compaction:
1) dealing with an edge case when a `TimedPut` entry's internal key is the right bound for penultimate level, the internal key after swapping in its preferred sequence number will fall outside of the penultimate range because preferred sequence number is smaller than its original sequence number. The entry however is still safe to be placed on penultimate level, so we keep track of `TimedPut` entry's original sequence number for this check. The idea behind this is that as long as it's safe for the original key to be placed on penultimate level, it's safe for the entry with swapped preferred sequence number to be placed on penultimate level too. Because we only swap in preferred sequence number when that entry is visible to the earliest snapshot and there is no other data points with the same user key in lower levels. On the other hand, as long as it's not safe for the original key to be placed on penultimate level, we will not place the entry after swapping the preferred seqno on penultimate level either.
2) the assertion that preferred seqno is always bigger than original sequence number may fail if this logic is only exercised after sequence number is zeroed out. We adjust the assertion to handle that case too. In this case, we don't swap in the preferred seqno but will adjust the its type to `kTypeValue`.
3) there was a special case handling for when range deletion may end up incorrectly covering an entry if preferred seqno is swapped in. But it missed the case that if the original entry is already covered by range deletion. The original handling will mistakenly output the entry instead of omitting it.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/12543
Test Plan:
./tiered_compaction_test --gtest_filter="PrecludeLastLevelTest.PreserveTimedPutOnPenultimateLevel"
./compaction_iterator_test --gtest_filter="*TimedPut*"
Reviewed By: pdillinger
Differential Revision: D56195096
Pulled By: jowlyzhang
fbshipit-source-id: 37ebb09d2513abbd9e90cda0217e26874584b8f3
Summary:
`check_flush_compaction_key_order` option was introduced for the key order checking online validation. It gave users the ability to disable the validation without downgrade in case the validation caused inefficiencies or false positives. Over time this validation has shown to be cheap and correct, so the option to disable it can now be removed.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/12311
Reviewed By: cbi42
Differential Revision: D53233379
Pulled By: ajkr
fbshipit-source-id: 1384361104021d6e3e580dce2ec123f9f99ce637
Summary:
This change is before a planned DBImpl change to ensure all sufficiently recent sequence numbers since Open are covered by SeqnoToTimeMapping (bug fix with existing test work-arounds). **Intended follow-up**
However, I found enough issues with SeqnoToTimeMapping to warrant this PR first, including very small fixes in DB implementation related to API contract of SeqnoToTimeMapping.
Functional fixes / changes:
* This fixes some mishandling of boundary cases. For example, if the user decides to stop writing to DB, the last written sequence number would perpetually have its write time updated to "now" and would always be ineligible for migration to cold tier. Part of the problem is that the SeqnoToTimeMapping would return a seqno known to have been written before (immediately or otherwise) the requested time, but compaction_job.cc would include that seqno in the preserve/exclude set. That is fixed (in part) by adding one in compaction_job.cc
* That problem was worse because a whole range of seqnos could be updated perpetually with new times in SeqnoToTimeMapping::Append (if no writes to DB). That logic was apparently optimized for GetOldestApproximateTime (now GetProximalTimeBeforeSeqno), which is not used in production, to the detriment of GetOldestSequenceNum (now GetProximalSeqnoBeforeTime), which is used in production. (Perhaps plans changed during development?) This is fixed in Append to optimize for accuracy of GetProximalSeqnoBeforeTime. (Unit tests added and updated.)
* Related: SeqnoToTimeMapping did not have a clear contract about the relationships between seqnos and times, just the idea of a rough correspondence. Now the class description makes it clear that the write time of each recorded seqno comes before or at the associated time, to support getting best results for GetProximalSeqnoBeforeTime. And this makes it easier to make clear the contract of each API function.
* Update `DBImpl::RecordSeqnoToTimeMapping()` to follow this ordering in gathering samples.
Some part of these changes has required an expanded test work-around for the problem (see intended follow-up above) that the DB does not immediately ensure recent seqnos are covered by its mapping. These work-arounds will be removed with that planned work.
An apparent compaction bug is revealed in
PrecludeLastLevelTest::RangeDelsCauseFileEndpointsToOverlap, so that test is disabled. Filed GitHub issue #11909
Cosmetic / code safety things (not exhaustive):
* Fix some confusing names.
* `seqno_time_mapping` was used inconsistently in places. Now just `seqno_to_time_mapping` to correspond to class name.
* Rename confusing `GetOldestSequenceNum` -> `GetProximalSeqnoBeforeTime` and `GetOldestApproximateTime` -> `GetProximalTimeBeforeSeqno`. Part of the motivation is that our times and seqnos here have the same underlying type, so we want to be clear about which is expected where to avoid mixing.
* Rename `kUnknownSeqnoTime` to `kUnknownTimeBeforeAll` because the value is a bad choice for unknown if we ever add ProximalAfterBlah functions.
* Arithmetic on SeqnoTimePair doesn't make sense except for delta encoding, so use better names / APIs with that in mind.
* (OMG) Don't allow direct comparison between SeqnoTimePair and SequenceNumber. (There is no checking that it isn't compared against time by accident.)
* A field name essentially matching the containing class name is a confusing pattern (`seqno_time_mapping_`).
* Wrap calls to confusing (but useful) upper_bound and lower_bound functions to have clearer names and more code reuse.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11905
Test Plan: GetOldestSequenceNum (now GetProximalSeqnoBeforeTime) and TruncateOldEntries were lacking unit tests, despite both being used in production (experimental feature). Added those and expanded others.
Reviewed By: jowlyzhang
Differential Revision: D49755592
Pulled By: pdillinger
fbshipit-source-id: f72a3baac74d24b963c77e538bba89a7fc8dce51
Summary:
Similar to point tombstones, we can drop a range tombstone during compaction when we know its range does not exist in any higher level. This PR adds this optimization. Some existing test in db_range_del_test is fixed to work under this optimization.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11459
Test Plan:
* Add unit test `DBRangeDelTest, NonBottommostCompactionDropRangetombstone`.
* Ran crash test that issues range deletion for a few hours: `python3 tools/db_crashtest.py blackbox --simple --write_buffer_size=1048576 --delrangepercent=10 --writepercent=31 --readpercent=40`
Reviewed By: ajkr
Differential Revision: D46007904
Pulled By: cbi42
fbshipit-source-id: 3f37205b6778b7d55ed106369ca41b0632a6d0fd
Summary:
A second attempt after https://github.com/facebook/rocksdb/issues/10802, with bug fixes and refactoring. This PR updates compaction logic to take range tombstones into account when determining whether to cut the current compaction output file (https://github.com/facebook/rocksdb/issues/4811). Before this change, only point keys were considered, and range tombstones could cause large compactions. For example, if the current compaction outputs is a range tombstone [a, b) and 2 point keys y, z, they would be added to the same file, and may overlap with too many files in the next level and cause a large compaction in the future. This PR also includes ajkr's effort to simplify the logic to add range tombstones to compaction output files in `AddRangeDels()` ([https://github.com/facebook/rocksdb/issues/11078](https://github.com/facebook/rocksdb/pull/11078#issuecomment-1386078861)).
The main change is for `CompactionIterator` to emit range tombstone start keys to be processed by `CompactionOutputs`. A new class `CompactionMergingIterator` is introduced to replace `MergingIterator` under `CompactionIterator` to enable emitting of range tombstone start keys. Further improvement after this PR include cutting compaction output at some grandparent boundary key (instead of the next output key) when cutting within a range tombstone to reduce overlap with grandparents.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11113
Test Plan:
* added unit test in db_range_del_test
* crash test with a small key range: `python3 tools/db_crashtest.py blackbox --simple --max_key=100 --interval=600 --write_buffer_size=262144 --target_file_size_base=256 --max_bytes_for_level_base=262144 --block_size=128 --value_size_mult=33 --subcompactions=10 --use_multiget=1 --delpercent=3 --delrangepercent=2 --verify_iterator_with_expected_state_one_in=2 --num_iterations=10`
Reviewed By: ajkr
Differential Revision: D42655709
Pulled By: cbi42
fbshipit-source-id: 8367e36ef5640e8f21c14a3855d4a8d6e360a34c
Summary:
in `CompactionOutputs::ShouldStopBefore()`, TTL-related states, `cur_files_to_cut_for_ttl_` and `next_files_to_cut_for_ttl_`, are not updated if the function returns early. This can cause unnecessary compaction output file cuttings and hence produce smaller output files, which may hurt write amp. See the example in the unit test for how this "unnecessary file cutting" can happen. This PR fixes this issue by moving the code for updating TTL states earlier in `CompactionOutputs::ShouldStopBefore()` so that the states are updated for each key.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11075
Test Plan: - Added new unit test.
Reviewed By: hx235
Differential Revision: D42398739
Pulled By: cbi42
fbshipit-source-id: 09fab66679c1a734abcfc31bcea33dd9aeb9dbc7
Summary:
This reverts commit f02c708aa3 since it introduced several bugs (see https://github.com/facebook/rocksdb/issues/11078 and https://github.com/facebook/rocksdb/issues/11067 for attempts to fix them) and that I do not have a high confidence to fix all of them and ensure no further ones before the next release branch cut. There are also come existing issue found during bug fixing. We will work on it and try to merge it to the release after.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11089
Test Plan: existing CI.
Reviewed By: ajkr
Differential Revision: D42505972
Pulled By: cbi42
fbshipit-source-id: 2f66dcde6b85dc94977b317c2ce513872cfbc153
Summary:
This PR is the first step for Issue https://github.com/facebook/rocksdb/issues/4811. Currently compaction output files are cut at point keys, and the decision is made mainly in `CompactionOutputs::ShouldStopBefore()`. This makes it possible for range tombstones to cause large compactions that does not respect `max_compaction_bytes`. For example, we can have a large range tombstone that overlaps with too many files from the next level. Another example is when there is a gap between a range tombstone and another key. The first issue may be more acceptable, as a lot of data is deleted. This PR address the second issue by calling `ShouldStopBefore()` for range tombstone start keys. The main change is for `CompactionIterator` to emit range tombstone start keys to be processed by `CompactionOutputs`. A new `CompactionMergingIterator` is introduced and only used under `CompactionIterator` for this purpose. Further improvement after this PR include 1) cut compaction output at some grandparent boundary key instead of at the next point key or range tombstone start key and 2) cut compaction output file within a large range tombstone (it may be easier and reasonable to only do it for range tombstones at the end of a compaction output).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/10802
Test Plan:
- added unit tests in db_range_del_test.
- stress test: `python3 tools/db_crashtest.py whitebox --[simple|enable_ts] --verify_iterator_with_expected_state_one_in=5 --delrangepercent=5 --prefixpercent=2 --writepercent=58 --readpercen=21 --duration=36000 --range_deletion_width=1000000`
Reviewed By: ajkr, jay-zhuang
Differential Revision: D40308827
Pulled By: cbi42
fbshipit-source-id: a8fd6f70a3f09d0ef7a40e006f6c964bba8c00df
Summary:
To make it consistent with the compaction picker which uses the `sstableKeyCompare()` to pick the overlap files. For example, without this change, it may cut L1 files like:
```
L1: [2-21] [22-30]
L2: [1-10] [21-30]
```
Because "21" on L1 is smaller than "21" on L2. But for compaction, these 2 files are overlapped.
`sstableKeyCompare()` also take range delete into consideration which may cut file for the same key.
It also makes the `max_compaction_bytes` calculation more accurate for cases like above, the overlapped bytes was under estimated. Also make sure the 2 keys won't be splitted to 2 files because of reaching `max_compaction_bytes`.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/10763
Reviewed By: cbi42
Differential Revision: D39971904
Pulled By: cbi42
fbshipit-source-id: bcc309e9c3dc61a8f50667a6f633e6132c0154a8
Summary:
Add user-defined timestamp support for range deletion. The new API is `DeleteRange(opt, cf, begin_key, end_key, ts)`. Most of the change is to update the comparator to compare without timestamp. Other than that, major changes are
- internal range tombstone data structures (`FragmentedRangeTombstoneList`, `RangeTombstone`, etc.) to store timestamps.
- Garbage collection of range tombstones and range tombstone covered keys during compaction.
- Get()/MultiGet() to return the timestamp of a range tombstone when needed.
- Get/Iterator with range tombstones bounded by readoptions.timestamp.
- timestamp crash test now issues DeleteRange by default.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/10661
Test Plan:
- Added unit test: `make check`
- Stress test: `python3 tools/db_crashtest.py --enable_ts whitebox --readpercent=57 --prefixpercent=4 --writepercent=25 -delpercent=5 --iterpercent=5 --delrangepercent=4`
- Ran `db_bench` to measure regression when timestamp is not enabled. The tests are for write (with some range deletion) and iterate with DB fitting in memory: `./db_bench--benchmarks=fillrandom,seekrandom --writes_per_range_tombstone=200 --max_write_buffer_number=100 --min_write_buffer_number_to_merge=100 --writes=500000 --reads=500000 --seek_nexts=10 --disable_auto_compactions -disable_wal=true --max_num_range_tombstones=1000`. Did not see consistent regression in no timestamp case.
| micros/op | fillrandom | seekrandom |
| --- | --- | --- |
|main| 2.58 |10.96|
|PR 10661| 2.68 |10.63|
Reviewed By: riversand963
Differential Revision: D39441192
Pulled By: cbi42
fbshipit-source-id: f05aca3c41605caf110daf0ff405919f300ddec2
Summary:
Try to align the compaction output file boundaries to the next level ones
(grandparent level), to reduce the level compaction write-amplification.
In level compaction, there are "wasted" data at the beginning and end of the
output level files. Align the file boundary can avoid such "wasted" compaction.
With this PR, it tries to align the non-bottommost level file boundaries to its
next level ones. It may cut file when the file size is large enough (at least
50% of target_file_size) and not too large (2x target_file_size).
db_bench shows about 12.56% compaction reduction:
```
TEST_TMPDIR=/data/dbbench2 ./db_bench --benchmarks=fillrandom,readrandom -max_background_jobs=12 -num=400000000 -target_file_size_base=33554432
# baseline:
Flush(GB): cumulative 25.882, interval 7.216
Cumulative compaction: 285.90 GB write, 162.36 MB/s write, 269.68 GB read, 153.15 MB/s read, 2926.7 seconds
# with this change:
Flush(GB): cumulative 25.882, interval 7.753
Cumulative compaction: 249.97 GB write, 141.96 MB/s write, 233.74 GB read, 132.74 MB/s read, 2534.9 seconds
```
The compaction simulator shows a similar result (14% with 100G random data).
As a side effect, with this PR, the SST file size can exceed the
target_file_size, but is capped at 2x target_file_size. And there will be
smaller files. Here are file size statistics when loading 100GB with the target
file size 32MB:
```
baseline this_PR
count 1.656000e+03 1.705000e+03
mean 3.116062e+07 3.028076e+07
std 7.145242e+06 8.046139e+06
```
The feature is enabled by default, to revert to the old behavior disable it
with `AdvancedColumnFamilyOptions.level_compaction_dynamic_file_size = false`
Also includes https://github.com/facebook/rocksdb/issues/1963 to cut file before skippable grandparent file. Which is for
use case like user adding 2 or more non-overlapping data range at the same
time, it can reduce the overlapping of 2 datasets in the lower levels.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/10655
Reviewed By: cbi42
Differential Revision: D39552321
Pulled By: jay-zhuang
fbshipit-source-id: 640d15f159ab0cd973f2426cfc3af266fc8bdde2
Summary:
Which will be used for tiered storage to preclude hot data from
compacting to the cold tier (the last level).
Internally, adding seqno to time mapping. A periodic_task is scheduled
to record the current_seqno -> current_time in certain cadence. When
memtable flush, the mapping informaiton is stored in sstable property.
During compaction, the mapping information are merged and get the
approximate time of sequence number, which is used to determine if a key
is recently inserted or not and preclude it from the last level if it's
recently inserted (within the `preclude_last_level_data_seconds`).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/10338
Test Plan: CI
Reviewed By: siying
Differential Revision: D37810187
Pulled By: jay-zhuang
fbshipit-source-id: 6953be7a18a99de8b1cb3b162d712f79c2b4899f
Summary:
Support per_key_placement for last level compaction, which will
be used for tiered compaction.
* compaction iterator reports which level a key should output to;
* compaction get the output level information and check if it's safe to
output the data to penultimate level;
* all compaction output files will be installed.
* extra internal compaction stats added for penultimate level.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/9964
Test Plan:
* Unittest
* db_bench, no significate difference: https://gist.github.com/jay-zhuang/3645f8fb97ec0ab47c10704bb39fd6e4
* microbench manual compaction no significate difference: https://gist.github.com/jay-zhuang/ba679b3e89e24992615ee9eef310e6dd
* run the db_stress multiple times (not covering the new feature) looks good (internal: https://fburl.com/sandcastle/9w84pp2m)
Reviewed By: ajkr
Differential Revision: D36249494
Pulled By: jay-zhuang
fbshipit-source-id: a96da57c8031c1df83e4a7a8567b657a112b80a3
Yu Zhang