3d57f4c961
[ #2179 ] test: Fix test TestEvacuateNetwork/multiple_shards,_evacuate_many
...
Signed-off-by: Anton Nikiforov <an.nikiforov@yadro.com>
2023-01-24 13:37:49 +03:00
9936b112b8
[ #5 ] blobstor: Use generic LRU cache
...
Signed-off-by: Evgenii Stratonikov <e.stratonikov@yadro.com>
2022-12-31 23:04:06 +03:00
4155c1bdff
[ #5 ] writecache: Use generic LRU cache
...
Signed-off-by: Evgenii Stratonikov <e.stratonikov@yadro.com>
2022-12-31 23:04:06 +03:00
0272218eb9
[ #2184 ] compression: Properly calculate upper bound
...
If the data is not compressible allocating `len(data)` will lead to a
slice reallocation. For a compressible data the results for small size
are flaky and we allocate a bit more. However, it feels right to use a
provided function if we need to pick any size at all.
```
name old time/op new time/op delta
Compression/size=128/zeroed_slice-8 2.23µs ±12% 2.06µs ± 6% -7.35% (p=0.009 n=10+10)
Compression/size=128/not_so_random_slice_(block_=_123)-8 19.0µs ±10% 15.8µs ±16% -17.09% (p=0.000 n=9+10)
Compression/size=128/random_slice-8 17.6µs ±15% 16.1µs ±16% ~ (p=0.075 n=10+10)
Compression/size=1024/zeroed_slice-8 3.05µs ±11% 2.84µs ±10% ~ (p=0.089 n=10+10)
Compression/size=1024/not_so_random_slice_(block_=_123)-8 18.1µs ± 6% 18.2µs ±12% ~ (p=0.971 n=10+10)
Compression/size=1024/random_slice-8 48.6µs ± 6% 45.6µs ± 5% -6.07% (p=0.006 n=10+9)
Compression/size=32768/zeroed_slice-8 26.8µs ± 3% 28.7µs ± 8% +7.23% (p=0.001 n=10+10)
Compression/size=32768/not_so_random_slice_(block_=_123)-8 44.3µs ± 8% 43.7µs ±13% ~ (p=0.762 n=8+10)
Compression/size=32768/random_slice-8 97.3µs ±32% 68.9µs ±15% -29.13% (p=0.000 n=10+10)
Compression/size=33554432/zeroed_slice-8 29.8ms ± 9% 30.3ms ±17% ~ (p=1.000 n=9+9)
Compression/size=33554432/not_so_random_slice_(block_=_123)-8 33.1ms ±14% 30.3ms ±11% -8.61% (p=0.043 n=10+10)
Compression/size=33554432/random_slice-8 41.7ms ± 3% 30.1ms ± 8% -27.72% (p=0.000 n=9+10)
name old alloc/op new alloc/op delta
Compression/size=128/zeroed_slice-8 128B ± 0% 144B ± 0% +12.50% (p=0.000 n=10+10)
Compression/size=128/not_so_random_slice_(block_=_123)-8 384B ± 0% 144B ± 0% -62.50% (p=0.000 n=10+10)
Compression/size=128/random_slice-8 384B ± 0% 144B ± 0% -62.50% (p=0.000 n=10+10)
Compression/size=1024/zeroed_slice-8 1.02kB ± 0% 1.15kB ± 0% +12.50% (p=0.000 n=10+10)
Compression/size=1024/not_so_random_slice_(block_=_123)-8 1.02kB ± 0% 1.15kB ± 0% +12.50% (p=0.000 n=10+10)
Compression/size=1024/random_slice-8 2.56kB ± 0% 1.15kB ± 0% -55.00% (p=0.000 n=10+10)
Compression/size=32768/zeroed_slice-8 32.8kB ± 0% 41.0kB ± 0% +25.00% (p=0.000 n=10+10)
Compression/size=32768/not_so_random_slice_(block_=_123)-8 32.8kB ± 0% 41.0kB ± 0% +25.00% (p=0.000 n=10+10)
Compression/size=32768/random_slice-8 81.9kB ± 0% 41.0kB ± 0% -50.00% (p=0.000 n=10+10)
Compression/size=33554432/zeroed_slice-8 33.6MB ± 0% 33.6MB ± 0% +0.02% (p=0.000 n=9+9)
Compression/size=33554432/not_so_random_slice_(block_=_123)-8 33.6MB ± 0% 33.6MB ± 0% +0.02% (p=0.000 n=8+10)
Compression/size=33554432/random_slice-8 75.5MB ± 0% 33.6MB ± 0% -55.55% (p=0.000 n=10+10)
name old allocs/op new allocs/op delta
Compression/size=128/zeroed_slice-8 1.00 ± 0% 1.00 ± 0% ~ (all equal)
Compression/size=128/not_so_random_slice_(block_=_123)-8 2.00 ± 0% 1.00 ± 0% -50.00% (p=0.000 n=10+10)
Compression/size=128/random_slice-8 2.00 ± 0% 1.00 ± 0% -50.00% (p=0.000 n=10+10)
Compression/size=1024/zeroed_slice-8 1.00 ± 0% 1.00 ± 0% ~ (all equal)
Compression/size=1024/not_so_random_slice_(block_=_123)-8 1.00 ± 0% 1.00 ± 0% ~ (all equal)
Compression/size=1024/random_slice-8 2.00 ± 0% 1.00 ± 0% -50.00% (p=0.000 n=10+10)
Compression/size=32768/zeroed_slice-8 1.00 ± 0% 1.00 ± 0% ~ (all equal)
Compression/size=32768/not_so_random_slice_(block_=_123)-8 1.00 ± 0% 1.00 ± 0% ~ (all equal)
Compression/size=32768/random_slice-8 2.00 ± 0% 1.00 ± 0% -50.00% (p=0.000 n=10+10)
Compression/size=33554432/zeroed_slice-8 1.00 ± 0% 1.00 ± 0% ~ (all equal)
Compression/size=33554432/not_so_random_slice_(block_=_123)-8 1.00 ± 0% 1.00 ± 0% ~ (all equal)
Compression/size=33554432/random_slice-8 2.00 ± 0% 1.00 ± 0% -50.00% (p=0.000 n=10+10)
```
Signed-off-by: Evgenii Stratonikov <e.stratonikov@yadro.com>
2022-12-30 11:07:35 +03:00
0ace28e43d
[ #2175 ] blobovniczatree: Close all non-active blobovniczas
...
Signed-off-by: Evgenii Stratonikov <e.stratonikov@yadro.com>
2022-12-30 11:07:35 +03:00
c1cf418956
[ #2175 ] blobovniczatree: Make function parameters more descriptive
...
Signed-off-by: Evgenii Stratonikov <e.stratonikov@yadro.com>
2022-12-30 11:07:35 +03:00
b4e90cdf51
[ #2165 ] pilorama: Optimize TreeApply when used for synchronization
...
Because synchronization _most likely_ will have apply already existing
operations, it is much faster to check their presence in a read
transaction. However, always doing this will degrade the perfomance
for normal `Apply`. And, let's be honest, it is already not good.
Thus we add a separate parameter which specifies whether this logic is
enabled.
Signed-off-by: Evgenii Stratonikov <e.stratonikov@yadro.com>
2022-12-30 11:07:35 +03:00
Pavel Karpy
21717262ec
[ #2016 ] shard: Check meta first on Get
...
`meta` should prevent returning removed objects (`GCMark` and `TS` relations
are `meta` abstractions).
Signed-off-by: Pavel Karpy <p.karpy@yadro.com>
2022-12-30 11:07:35 +03:00
Pavel Karpy
74ec71446f
[ #2167 ] shard: Do not use write-cache by default in Head
...
Both `meta` and `write-cache` are expected to have a fast underlying disk,
so it does not seem like an optimisation. Moreover, `write-cache`'s `Head`
is a `Get` with payload cutting, it _must_ use more memory for no reason
(`meta` was created for such requests). Also, `write-cache` does not allow
performing any "meta" relations checks (such as locking, tombstoning).
Signed-off-by: Pavel Karpy <p.karpy@yadro.com>
2022-12-30 11:07:35 +03:00
Pavel Karpy
1608fd1c07
[ #2167 ] write-cache: Add "write-cache" to its logs
...
Signed-off-by: Pavel Karpy <p.karpy@yadro.com>
2022-12-30 11:07:35 +03:00
Pavel Karpy
eea2892109
[ #1956 ] node: Lock shard's mode on its methods switch
...
Signed-off-by: Pavel Karpy <p.karpy@yadro.com>
2022-12-30 11:07:35 +03:00
Evgenii Stratonikov
e1c3bdbfa6
[ #1621 ] pilorama: Remove Timestamp field from nodeInfo
...
It is already present in `Meta`.
Signed-off-by: Evgenii Stratonikov <evgeniy@nspcc.ru>
Signed-off-by: Evgenii Stratonikov <evgeniy@morphbits.ru>
2022-12-30 11:07:35 +03:00
Evgenii Stratonikov
1044adbe94
[ #1621 ] pilorama: Improve memory allocation
...
Signed-off-by: Evgenii Stratonikov <evgeniy@nspcc.ru>
Signed-off-by: Evgenii Stratonikov <evgeniy@morphbits.ru>
2022-12-30 11:07:35 +03:00
Evgenii Stratonikov
2539d466a6
[ #1621 ] pilorama: Seek after cursor invalidation
...
Signed-off-by: Evgenii Stratonikov <evgeniy@nspcc.ru>
Signed-off-by: Evgenii Stratonikov <evgeniy@morphbits.ru>
2022-12-30 11:07:35 +03:00
Evgenii Stratonikov
e9ba8931f8
[ #1621 ] pilorama: Simplify bucket creation
...
Signed-off-by: Evgenii Stratonikov <evgeniy@nspcc.ru>
Signed-off-by: Evgenii Stratonikov <evgeniy@morphbits.ru>
2022-12-30 11:07:35 +03:00
Evgenii Stratonikov
fe7ddfdc6a
[ #1621 ] pilorama: Compare memory forests properly
...
Node children are not sorted and could occur in any order.
Signed-off-by: Evgenii Stratonikov <evgeniy@nspcc.ru>
Signed-off-by: Evgenii Stratonikov <evgeniy@morphbits.ru>
2022-12-30 11:07:35 +03:00
edb1428248
[ #2022 ] Add metric readonly to get shards mode
...
Signed-off-by: Anton Nikiforov <an.nikiforov@yadro.com>
2022-12-30 11:07:35 +03:00
e5c304536b
[ #2161 ] pilorama: Do not apply already existing operations
...
Speeds up synchronization a bit.
Signed-off-by: Evgenii Stratonikov <e.stratonikov@yadro.com>
2022-12-30 11:07:35 +03:00
Pavel Karpy
21c58c92a9
[ #2145 ] meta: Do allow force inhuming a locked object
...
Signed-off-by: Pavel Karpy <p.karpy@yadro.com>
2022-12-30 11:07:35 +03:00
0b78af467e
[ #2140 ] engine: Fix error handling in TreeMove
...
Signed-off-by: Evgenii Stratonikov <e.stratonikov@yadro.com>
2022-12-30 11:07:35 +03:00
Pavel Karpy
923f84722a
Move to frostfs-node
...
Signed-off-by: Pavel Karpy <p.karpy@yadro.com>
2022-12-28 15:04:29 +03:00
Evgenii Stratonikov
42554a9298
[ #2068 ] writecache: Remove deleted objects from the writecache
...
Signed-off-by: Evgenii Stratonikov <evgeniy@morphbits.ru>
2022-12-02 11:52:05 +03:00
Evgenii Stratonikov
4a49ea0855
[ #2068 ] writecache: Allow to open FSTree in read-only mode
...
Signed-off-by: Evgenii Stratonikov <evgeniy@morphbits.ru>
2022-12-02 11:52:05 +03:00
Evgenii Stratonikov
857d2dc3f5
[ #2068 ] writecache: Optimize initial flush existence checking
...
Signed-off-by: Evgenii Stratonikov <evgeniy@morphbits.ru>
2022-12-02 11:52:05 +03:00
Evgenii Stratonikov
63f604e948
[ #2068 ] blobstor: Allow to provide storage ID in Exists
...
Signed-off-by: Evgenii Stratonikov <evgeniy@morphbits.ru>
2022-12-02 11:52:05 +03:00
Evgenii Stratonikov
6ad2b5d5b8
[ #2068 ] blobovnicza: Add Exists method
...
Signed-off-by: Evgenii Stratonikov <evgeniy@morphbits.ru>
2022-12-02 11:52:05 +03:00
Pavel Karpy
3d0768a1d3
[ #2061 ] node: Unify meta.Get benchmarks
...
Make them get exactly one (different) object per a bench iteration.
Signed-off-by: Pavel Karpy <carpawell@nspcc.ru>
2022-11-30 18:29:14 +03:00
Pavel Karpy
bc905f169d
[ #2061 ] meta: Add parallel bench for Get
...
Signed-off-by: Pavel Karpy <carpawell@nspcc.ru>
2022-11-30 18:29:14 +03:00
Evgenii Stratonikov
7335a52f29
[ #1732 ] pilorama: Improve logical error handling
...
Signed-off-by: Evgenii Stratonikov <evgeniy@morphbits.ru>
2022-11-30 16:58:52 +03:00
Evgenii Stratonikov
ae7b473768
[ #2064 ] blobovniczatree: Remove index too big log
...
There is no need to log about a situation which is expected.
Signed-off-by: Evgenii Stratonikov <evgeniy@morphbits.ru>
2022-11-30 16:53:18 +03:00
Pavel Karpy
ed4351aab0
[ #2074 ] write-cache: Do not flush same object twice
...
Signed-off-by: Pavel Karpy <carpawell@nspcc.ru>
2022-11-19 11:01:04 +03:00
Pavel Karpy
dd225906a0
[ #2074 ] write-cache: Remove unused variables
...
Signed-off-by: Pavel Karpy <carpawell@nspcc.ru>
2022-11-19 11:01:04 +03:00
Pavel Karpy
b1025bdb42
[ #2057 ] meta: Fail write operations in R/O mode
...
Signed-off-by: Pavel Karpy <carpawell@nspcc.ru>
2022-11-19 11:01:04 +03:00
Pavel Karpy
fdeea1dfac
[ #2057 ] meta: Fix concurrent mode changes
...
Includes:
1. mode change read lock operation in every exported method that r/w the
underlying database;
2. returning `ErrDegradedMode` logical error if any exported method is
called in degraded (without a metabase) mode.
Signed-off-by: Pavel Karpy <carpawell@nspcc.ru>
2022-11-19 11:01:04 +03:00
Pavel Karpy
3d6defd3e8
[ #2057 ] meta: Do not lock the whole meta on GET
...
Signed-off-by: Pavel Karpy <carpawell@nspcc.ru>
2022-11-19 11:01:04 +03:00
Pavel Karpy
fa231b8c56
[ #2057 ] blobstor: Block operations on a mode change
...
Signed-off-by: Pavel Karpy <carpawell@nspcc.ru>
2022-11-19 11:01:04 +03:00
Pavel Karpy
b673d9e472
[ #2053 ] engine: Do not switch mode because of logical errors
...
Signed-off-by: Pavel Karpy <carpawell@nspcc.ru>
2022-11-19 11:01:04 +03:00
9a20498f34
[ #1940 ] Removing all trees by container ID if tree ID is empty in pilorama.Forest.TreeDrop
...
Signed-off-by: Anton Nikiforov <an.nikiforov@yadro.com>
2022-11-19 11:01:04 +03:00
Pavel Karpy
634792077e
[ #1502 ] node: Store lock object on every container node
...
Includes extending listing methods in the Storage Engine with object types.
It allows tuning replication/policer algorithms: container nodes do
not remove `LOCK` objects as redundant and try to fulfill `LOCK` placement
on the ohter container nodes.
Signed-off-by: Pavel Karpy <carpawell@nspcc.ru>
2022-11-19 11:01:04 +03:00
Pavel Karpy
3b61cb4f49
[ #1502 ] engine: Check all shards for LOCK'ing before inhuming
...
It allows keeping all the locked objects safe after metabase
resynchronization. Currently, all `LOCK` objects are broadcast to all nodes
in a container, it guarantees `LOCK` object presence in a regular situation.
Signed-off-by: Pavel Karpy <carpawell@nspcc.ru>
2022-11-19 11:01:04 +03:00
Pavel Karpy
34e8d2ba56
[ #1502 ] shard: Add IsLocked method
...
Signed-off-by: Pavel Karpy <carpawell@nspcc.ru>
2022-11-19 11:01:04 +03:00
Pavel Karpy
9a039ba582
[ #1502 ] meta: Add IsLocked method
...
It gets an object and returns its locking status.
Signed-off-by: Pavel Karpy <carpawell@nspcc.ru>
2022-11-19 11:01:04 +03:00
Evgenii Stratonikov
d65604ad30
[ #1985 ] blobstor: Allow to report multiple errors to caller
...
Signed-off-by: Evgenii Stratonikov <evgeniy@morphbits.ru>
2022-11-19 11:01:04 +03:00
Evgenii Stratonikov
f2d7e65e39
[ #2035 ] engine: Allow moving to degraded from background workers
...
Signed-off-by: Evgenii Stratonikov <evgeniy@morphbits.ru>
2022-11-19 11:01:04 +03:00
Evgenii Stratonikov
b0e94b6a6b
[ #1906 ] writecache: Do not require read-only mode in Flush
...
It was needed before we started to flush during transition to
`degraded` mode. Now it is confusing.
Signed-off-by: Evgenii Stratonikov <evgeniy@morphbits.ru>
2022-11-19 11:01:04 +03:00
Pavel Karpy
2849e465f9
[ #1699 ] meta: Do not return SplitInfoError on Delete
...
It is not an error: removing virtual object is expected and should be just
skipped. Getting a virtual object with `raw` flag is considered as an
impossible action, all the virtual objects removals will be handled via
their children's removals implicitly.
Signed-off-by: Pavel Karpy <carpawell@nspcc.ru>
2022-11-19 11:01:04 +03:00
Evgenii Stratonikov
a3e7365cbd
[ #1732 ] pilorama: Fill parent mark correctly
...
Signed-off-by: Evgenii Stratonikov <evgeniy@morphbits.ru>
2022-11-19 11:01:04 +03:00
Evgenii Stratonikov
134f2ba02e
[ #1732 ] pilorama: Fix backwards log insertion
...
Signed-off-by: Evgenii Stratonikov <evgeniy@morphbits.ru>
2022-11-19 11:01:04 +03:00
Evgenii Stratonikov
2ef38cfbc4
[ #1996 ] engine: Ignore pilorama.ErrTreeNotFound for write operations
...
Signed-off-by: Evgenii Stratonikov <evgeniy@morphbits.ru>
2022-11-19 11:01:04 +03:00
Evgenii Stratonikov
d8d3588e1b
[ #1996 ] engine: Always select proper shard for a tree
...
Currently there is a possibility for modifying operations to fail
because of I/O errors and a new tree to be created on another shard.
This commit adds existence check for modifying operations.
Read operations remain as they are, not to slow things.
`TreeDrop` is an exception, because this is a tree removal and trying
multiple shards is not an unwanted behaviour.
Signed-off-by: Evgenii Stratonikov <evgeniy@morphbits.ru>
2022-11-03 15:29:23 +03:00
