Summary:
On MacOS with clang the compilation of _tools/db_bench_tool.cc_ always fails because the format used in a `fprintf` call has the wrong type. This PR should hopefully fix this issue
```
tools/db_bench_tool.cc:4233:61: error: format specifies type 'unsigned long long' but the argument has type 'size_t' (aka 'unsigned long')
```
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4533
Differential Revision: D10471657
Pulled By: maysamyabandeh
fbshipit-source-id: f20f5f3756d3571b586c895c845d0d4d1e34a398
Summary:
The new flag allows tombstones to be generated after enough
keys have been written to the database, which makes it easier to ensure
that tombstones cover a lot of keys.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4538
Differential Revision: D10455685
Pulled By: abhimadan
fbshipit-source-id: f25d5421745a353c830dea12b79784e852056551
Summary:
Current implementation of perf context is level agnostic. Making it hard to do performance evaluation for the LSM tree. This PR adds `PerfContextByLevel` to decompose the counters by level.
This will be helpful when analyzing point and range query performance as well as tuning bloom filter
Also replaced __thread with thread_local keyword for perf_context
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4226
Differential Revision: D10369509
Pulled By: miasantreble
fbshipit-source-id: f1ced4e0de5fcebdb7f9cff36164516bc6382d82
Summary:
This is a conceptually simple change, but it touches many files to
pass the allocator through function calls.
We introduce CacheAllocator, which can be used by clients to configure
custom allocator for cache blocks. Our motivation is to hook this up
with folly's `JemallocNodumpAllocator`
(f43ce6d686/folly/experimental/JemallocNodumpAllocator.h),
but there are many other possible use cases.
Additionally, this commit cleans up memory allocation in
`util/compression.h`, making sure that all allocations are wrapped in a
unique_ptr as soon as possible.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4437
Differential Revision: D10132814
Pulled By: yiwu-arbug
fbshipit-source-id: be1343a4b69f6048df127939fea9bbc96969f564
Summary:
If range tombstones are generated every few writes, the
KeyGenerator's limit is now extended to account for the additional
Next() calls. This is primarily important for `filluniquerandom`
benchmarks that enforce the call limit.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4404
Differential Revision: D9949326
Pulled By: abhimadan
fbshipit-source-id: 0bdfeb2cad2098dc0b8b029236dab5e4bef25e38
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
Summary:
Reverting is needed to unblock a user building against master, who is blocked for multiple days due to a thread-safety issue in `GetEmptyDict`. We haven't been able to fix it quickly, so reverting.
Simply ran `git revert 6c40806e51a89386d2b066fddf73d3fd03a36f65`. There were no merge conflicts.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4347
Differential Revision: D9668365
Pulled By: ajkr
fbshipit-source-id: 0c56334f0a23cf5ee0233d4e4679eae6709739cd
Summary:
In RocksDB, for a given SST file, all data blocks are compressed with the same dictionary. When we compress a block using the dictionary's raw bytes, the compression library first has to digest the dictionary to get it into a usable form. This digestion work is redundant and ideally should be done once per file.
ZSTD offers APIs for the caller to create and reuse a digested dictionary object (`ZSTD_CDict`). In this PR, we call `ZSTD_createCDict` once per file to digest the raw bytes. Then we use `ZSTD_compress_usingCDict` to compress each data block using the pre-digested dictionary. Once the file's created `ZSTD_freeCDict` releases the resources held by the digested dictionary.
There are a couple other changes included in this PR:
- Changed the parameter object for (un)compression functions from `CompressionContext`/`UncompressionContext` to `CompressionInfo`/`UncompressionInfo`. This avoids the previous pattern, where `CompressionContext`/`UncompressionContext` had to be mutated before calling a (un)compression function depending on whether dictionary should be used. I felt that mutation was error-prone so eliminated it.
- Added support for digested uncompression dictionaries (`ZSTD_DDict`) as well. However, this PR does not support reusing them across uncompression calls for the same file. That work is deferred to a later PR when we will store the `ZSTD_DDict` objects in block cache.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4251
Differential Revision: D9257078
Pulled By: ajkr
fbshipit-source-id: 21b8cb6bbdd48e459f1c62343780ab66c0a64438
Summary:
Add `--data_block_index_type` and `--data_block_hash_table_util_ratio` option to `db_bench`.
`--data_block_index_type` can be either of `binary` (default) or `binary_and_hash`;
`--data_block_hash_table_util_ratio` will be a double. The default value is `0.75`.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4281
Differential Revision: D9361476
Pulled By: fgwu
fbshipit-source-id: dc53e01acef9db81b9eec5e8a96f3bc8ed718c10
Summary:
db_bench's previous default compression level (-1) was not the default compression level in all libraries. In particular, in ZSTD negative values are valid compression levels, while ZSTD's default compression level is three.
This PR changes db_bench's default to be RocksDB's library-independent default compression level (see #3895). I also changed a couple other flags to get their default values from an options object directly rather than hardcoding.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4248
Differential Revision: D9235140
Pulled By: ajkr
fbshipit-source-id: be4e0722d59fa1968832183db36d1d20fcf11e5b
Summary:
A framework for tracing and replaying RocksDB operations.
A binary trace file is created by capturing the DB operations, and it can be replayed back at the same rate using db_bench.
- Column-families are supported
- Multi-threaded tracing is supported.
- TraceReader and TraceWriter are exposed to the user, so that tracing to various destinations can be enabled (say, to other messaging/logging services). By default, a FileTraceReader and FileTraceWriter are implemented to capture to a file and replay from it.
- This is not yet ideal to be enabled in production due to large performance overhead, but it can be safely tried out in a shadow setup, say, for analyzing RocksDB operations.
Currently supported DB operations:
- Writes:
-- Put
-- Merge
-- Delete
-- SingleDelete
-- DeleteRange
-- Write
- Reads:
-- Get (point lookups)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/3837
Differential Revision: D7974837
Pulled By: sagar0
fbshipit-source-id: 8ec65aaf336504bc1f6ed0feae67f6ed5ef97a72
Summary:
When running the tracing and analyzing, I found that MergeRandom benchmark in db_bench only access the default column family even the -num_column_families is specified > 1.
changes: Using the db_with_cfh as DB to randomly select the column family to execute the Merge operation if -num_column_families is specified > 1.
Tested with make asan_check and verified in tracing
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4155
Differential Revision: D8907888
Pulled By: zhichao-cao
fbshipit-source-id: 2b4bc8fe0e99c8f262f5be6b986c7025d62cf850
Summary:
Lint is not happy with some new code recently committed. Format them.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4161
Differential Revision: D8940582
Pulled By: siying
fbshipit-source-id: c9b43b1ef8c88b5e923911058b44eb77234b36b7
Summary:
Adding the string "PERF_CONTEXT:" before the perf_context stats are printed. Setting the filter policy if it's a block based table even when options are being loaded from the provided FLAGS_options_file.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4153
Differential Revision: D8905517
Pulled By: poojam23
fbshipit-source-id: 5956ed7882d39ec8ae654d5dadeb88727a36f0dd
Summary:
The patch makes sure that two parallel test threads will operate on different db paths. This enables using open source tools such as gtest-parallel to run the tests of a file in parallel.
Example: ``` ~/gtest-parallel/gtest-parallel ./table_test```
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4135
Differential Revision: D8846653
Pulled By: maysamyabandeh
fbshipit-source-id: 799bad1abb260e3d346bcb680d2ae207a852ba84
Summary:
Right now there is no support for enabling compaction filter in db_bench, we should add support for that to facilitate testing of compaction filter.
This PR adds a compaction filter called KeepFilter and make `Filter` always returns false, essentially a noop compaction filter. This will allow us to test compaction filter code path without having to support arbitrary compaction filters
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4106
Differential Revision: D8828517
Pulled By: miasantreble
fbshipit-source-id: 9ad76d04103eaa9d00da98334b4a39e542d26c41
Summary:
`DEFINE_uint32` was unavailable on some platforms, e.g., https://travis-ci.org/facebook/rocksdb/jobs/403352902. Use `DEFINE_uint64` instead which should work as it's used many times elsewhere in this file.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4129
Differential Revision: D8830311
Pulled By: ajkr
fbshipit-source-id: b4fc90ba3f50e649c070ce8069c68e530d731f05
Summary:
give control of how often stats are printed, including jemalloc stats if enabled. Previously the default was 10 minutes so we'd only see updated stats for very long benchmark runs.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4109
Differential Revision: D8796444
Pulled By: ajkr
fbshipit-source-id: fd7902fe3f105fae89322c4ab63316bba4a2b15e
Summary:
Top-level index in partitioned index/filter blocks are small and could be pinned in memory. So far we use that by cache_index_and_filter_blocks to false. This however make it difficult to keep account of the total memory usage. This patch introduces pin_top_level_index_and_filter which in combination with cache_index_and_filter_blocks=true keeps the top-level index in cache and yet pinned them to avoid cache misses and also cache lookup overhead.
Closes https://github.com/facebook/rocksdb/pull/4037
Differential Revision: D8596218
Pulled By: maysamyabandeh
fbshipit-source-id: 3a5f7f9ca6b4b525b03ff6bd82354881ae974ad2
Summary:
PR https://github.com/facebook/rocksdb/pull/3838 made some changes that triggers lint warnings.
Run `make format` to fix formatting as suggested by siying .
Also piggyback two changes:
1) fix singleton destruction order for windows and posix env
2) fix two clang warnings
Closes https://github.com/facebook/rocksdb/pull/3954
Differential Revision: D8272041
Pulled By: miasantreble
fbshipit-source-id: 7c4fd12bd17aac13534520de0c733328aa3c6c9f
Summary:
format_version=3 changes the format of SST index. This is however not being tested currently since tests only work with the default format_version which is currently 2. The patch extends the most related tests to also test for format_version=3.
Closes https://github.com/facebook/rocksdb/pull/3942
Differential Revision: D8238413
Pulled By: maysamyabandeh
fbshipit-source-id: 915725f55753dd8e9188e802bf471c23645ad035
Summary:
Windows does not have LD_PRELOAD mechanism to override all memory allocation functions and ZSTD makes use of C-tuntime calloc. During flushes and compactions default system allocator fragments and the system slows down considerably.
For builds with jemalloc we employ an advanced ZSTD context creation API that re-directs memory allocation to jemalloc. To reduce the cost of context creation on each block we cache ZSTD context within the block based table builder while a new SST file is being built, this will help all platform builds including those w/o jemalloc. This avoids system allocator fragmentation and improves the performance.
The change does not address random reads and currently on Windows reads with ZSTD regress as compared with SNAPPY compression.
Closes https://github.com/facebook/rocksdb/pull/3838
Differential Revision: D8229794
Pulled By: miasantreble
fbshipit-source-id: 719b622ab7bf4109819bc44f45ec66f0dd3ee80d
Summary:
Small format error below causes build to fail. I believe that this :
```
fprintf(stderr, "num reads to do %lu\n", reads_);
```
Can be changed to this:
```
fprintf(stderr, "num reads to do %" PRIu64 "\n", reads_);
```
Successful build
```
CC utilities/blob_db/blob_dump_tool.o
AR librocksdb_debug.a
ar: creating archive librocksdb_debug.a
/Applications/Xcode.app/Contents/Developer/Toolchains/XcodeDefault.xctoolchain/usr/bin/ranlib: file: librocksdb_debug.a(rocks_lua_compaction_filter.o) has no symbols
CC tools/db_bench.o
CC tools/db_bench_tool.o
tools/db_bench_tool.cc:4532:46: error: format specifies type 'unsigned long' but the argument has type 'int64_t' (aka 'long long') [-Werror,-Wformat]
fprintf(stderr, "num reads to do %lu\n", reads_);
~~~ ^~~~~~
%lld
1 error generated.
make: *** [tools/db_bench_tool.o] Error 1
```
```
$ cd rocksdb
$ make all
$ g++ --version
Configured with: --prefix=/Applications/Xcode.app/Contents/Developer/usr --with-gxx-include-dir=/usr/include/c++/4.2.1
Apple LLVM version 9.1.0 (clang-902.0.39.1)
Target: x86_64-apple-darwin17.5.0
Thread model: posix
InstalledDir: /Applications/Xcode.app/Contents/Developer/Toolchains/XcodeDefault.xctoolchain/usr/bin
```
Closes https://github.com/facebook/rocksdb/pull/3909
Differential Revision: D8215710
Pulled By: siying
fbshipit-source-id: 15e49fb02a818fec846e9f9b2a50e372b6b67751
Summary:
Implement midpoint insertion strategy where new blocks will be insert to the middle of LRU list, then move the head on the first hit in cache.
Closes https://github.com/facebook/rocksdb/pull/3877
Differential Revision: D8100895
Pulled By: yiwu-arbug
fbshipit-source-id: f4bd83cb8be469e5d02072cfc8bd66011391f3da
Summary:
Previously `DBOptions::use_direct_io_for_flush_and_compaction=true` combined with `DBOptions::use_direct_reads=false` could cause RocksDB to simultaneously read from two file descriptors for the same file, where background reads used direct I/O and foreground reads used buffered I/O. Our measurements found this mixed-mode I/O negatively impacted foreground read perf, compared to when only buffered I/O was used.
This PR makes the mixed-mode I/O situation impossible by repurposing `DBOptions::use_direct_io_for_flush_and_compaction` to only apply to background writes, and `DBOptions::use_direct_reads` to apply to all reads. There is no risk of direct background direct writes happening simultaneously with buffered reads since we never read from and write to the same file simultaneously.
Closes https://github.com/facebook/rocksdb/pull/3829
Differential Revision: D7915443
Pulled By: ajkr
fbshipit-source-id: 78bcbf276449b7e7766ab6b0db246f789fb1b279
Summary:
Background activities like compaction can negatively affect
latency of higher-priority tasks like request processing. To avoid this,
rocksdb already lowers the IO priority of background threads on Linux
systems. While this takes care of typical IO-bound systems, it does not
help much when CPU (temporarily) becomes the bottleneck. This is
especially likely when using more expensive compression settings.
This patch adds an API to allow for lowering the CPU priority of
background threads, modeled on the IO priority API. Benchmarks (see
below) show significant latency and throughput improvements when CPU
bound. As a result, workloads with some CPU usage bursts should benefit
from lower latencies at a given utilization, or should be able to push
utilization higher at a given request latency target.
A useful side effect is that compaction CPU usage is now easily visible
in common tools, allowing for an easier estimation of the contribution
of compaction vs. request processing threads.
As with IO priority, the implementation is limited to Linux, degrading
to a no-op on other systems.
Closes https://github.com/facebook/rocksdb/pull/3763
Differential Revision: D7740096
Pulled By: gwicke
fbshipit-source-id: e5d32373e8dc403a7b0c2227023f9ce4f22b413c
Summary:
this PR fixes a few failed contbuild:
1. ASAN memory leak in Block::NewIterator (table/block.cc:429). the proper destruction of first_level_iter_ and second_level_iter_ of two_level_iterator.cc is missing from the code after the refactoring in https://github.com/facebook/rocksdb/pull/3406
2. various unused param errors introduced by https://github.com/facebook/rocksdb/pull/3662
3. updated comment for `ForceReleaseCachedEntry` to emphasize the use of `force_erase` flag.
Closes https://github.com/facebook/rocksdb/pull/3718
Reviewed By: maysamyabandeh
Differential Revision: D7621192
Pulled By: miasantreble
fbshipit-source-id: 476c94264083a0730ded957c29de7807e4f5b146
Summary:
Provide a block_align option in BlockBasedTableOptions to allow
alignment of SST file data blocks. This will avoid higher
IOPS/throughput load due to < 4KB data blocks spanning 2 4KB pages.
When this option is set to true, the block alignment is set to lower of
block size and 4KB.
Closes https://github.com/facebook/rocksdb/pull/3502
Differential Revision: D7400897
Pulled By: anand1976
fbshipit-source-id: 04cc3bd144e88e3431a4f97604e63ad7a0f06d44
Summary:
Improving blob db FIFO eviction with the following changes,
* Change blob_dir_size to max_db_size. Take into account SST file size when computing DB size.
* FIFO now only take into account live sst files and live blob files. It is normal for disk usage to go over max_db_size because there are obsolete sst files and blob files pending deletion.
* FIFO eviction now also evict TTL blob files that's still open. It doesn't evict non-TTL blob files.
* If FIFO is triggered, it will pass an expiration and the current sequence number to compaction filter. Compaction filter will then filter inlined keys to evict those with an earlier expiration and smaller sequence number. So call LSM FIFO.
* Compaction filter also filter those blob indexes where corresponding blob file is gone.
* Add an event listener to listen compaction/flush event and update sst file size.
* Implement DB::Close() to make sure base db, as well as event listener and compaction filter, destruct before blob db.
* More blob db statistics around FIFO.
* Fix some locking issue when accessing a blob file.
Closes https://github.com/facebook/rocksdb/pull/3556
Differential Revision: D7139328
Pulled By: yiwu-arbug
fbshipit-source-id: ea5edb07b33dfceacb2682f4789bea61de28bbfa
Summary:
Some workloads (like my current benchmarking) may want partitioned indexes without partitioned filters. Particularly, when `-optimize_filters_for_hits=true`, the total index size may be larger than the total filter size, so it can make sense to hold all filters in-memory but not all indexes.
Closes https://github.com/facebook/rocksdb/pull/3492
Differential Revision: D6970092
Pulled By: ajkr
fbshipit-source-id: b7fa1828e1d13829339aefb90fd56eb7c5337f61
Summary:
This is to avoid run time error. Fail the db_bench immediately if cuckoo table is used but mmap_read is not specified.
Closes https://github.com/facebook/rocksdb/pull/3420
Differential Revision: D6838284
Pulled By: siying
fbshipit-source-id: 20893fa28d40fadc31e4ff154bed02f5a1bad341
Summary:
It failed every time. I guess people usually ran with assertions disabled.
Closes https://github.com/facebook/rocksdb/pull/3422
Differential Revision: D6822984
Pulled By: ajkr
fbshipit-source-id: 2e90db75618b26ac1c46ddfa9e03c095c7bf16e3
Summary:
The macro was added by mistake in #2372
Closes https://github.com/facebook/rocksdb/pull/3343
Differential Revision: D6681356
Pulled By: yiwu-arbug
fbshipit-source-id: 4180172fb0eaef4189c07f219241e0c261c03461
Summary:
**# Summary**
RocksDB uses SSE crc32 intrinsics to calculate the crc32 values but it does it in single way fashion (not pipelined on single CPU core). Intel's whitepaper () published an algorithm that uses 3-way pipelining for the crc32 intrinsics, then use pclmulqdq intrinsic to combine the values. Because pclmulqdq has overhead on its own, this algorithm will show perf gains on buffers larger than 216 bytes, which makes RocksDB a perfect user, since most of the buffers RocksDB call crc32c on is over 4KB. Initial db_bench show tremendous CPU gain.
This change uses the 3-way SSE algorithm by default. The old SSE algorithm is now behind a compiler tag NO_THREEWAY_CRC32C. If user compiles the code with NO_THREEWAY_CRC32C=1 then the old SSE Crc32c algorithm would be used. If the server does not have SSE4.2 at the run time the slow way (Non SSE) will be used.
**# Performance Test Results**
We ran the FillRandom and ReadRandom benchmarks in db_bench. ReadRandom is the point of interest here since it calculates the CRC32 for the in-mem buffers. We did 3 runs for each algorithm.
Before this change the CRC32 value computation takes about 11.5% of total CPU cost, and with the new 3-way algorithm it reduced to around 4.5%. The overall throughput also improved from 25.53MB/s to 27.63MB/s.
1) ReadRandom in db_bench overall metrics
PER RUN
Algorithm | run | micros/op | ops/sec |Throughput (MB/s)
3-way | 1 | 4.143 | 241387 | 26.7
3-way | 2 | 3.775 | 264872 | 29.3
3-way | 3 | 4.116 | 242929 | 26.9
FastCrc32c|1 | 4.037 | 247727 | 27.4
FastCrc32c|2 | 4.648 | 215166 | 23.8
FastCrc32c|3 | 4.352 | 229799 | 25.4
AVG
Algorithm | Average of micros/op | Average of ops/sec | Average of Throughput (MB/s)
3-way | 4.01 | 249,729 | 27.63
FastCrc32c | 4.35 | 230,897 | 25.53
2) Crc32c computation CPU cost (inclusive samples percentage)
PER RUN
Implementation | run | TotalSamples | Crc32c percentage
3-way | 1 | 4,572,250,000 | 4.37%
3-way | 2 | 3,779,250,000 | 4.62%
3-way | 3 | 4,129,500,000 | 4.48%
FastCrc32c | 1 | 4,663,500,000 | 11.24%
FastCrc32c | 2 | 4,047,500,000 | 12.34%
FastCrc32c | 3 | 4,366,750,000 | 11.68%
**# Test Plan**
make -j64 corruption_test && ./corruption_test
By default it uses 3-way SSE algorithm
NO_THREEWAY_CRC32C=1 make -j64 corruption_test && ./corruption_test
make clean && DEBUG_LEVEL=0 make -j64 db_bench
make clean && DEBUG_LEVEL=0 NO_THREEWAY_CRC32C=1 make -j64 db_bench
Closes https://github.com/facebook/rocksdb/pull/3173
Differential Revision: D6330882
Pulled By: yingsu00
fbshipit-source-id: 8ec3d89719533b63b536a736663ca6f0dd4482e9
Summary:
I started adding gflags support for cmake on linux and got frustrated that I'd need to duplicate the build_detect_platform logic, which determines namespace based on attempting compilation. We can do it differently -- use the GFLAGS_NAMESPACE macro if available, and if not, that indicates it's an old gflags version without configurable namespace so we can simply hardcode "google".
Closes https://github.com/facebook/rocksdb/pull/3212
Differential Revision: D6456973
Pulled By: ajkr
fbshipit-source-id: 3e6d5bde3ca00d4496a120a7caf4687399f5d656
Simon Grätzer