Each Put, List, Multipart operations heavily rely on making
GetBucketInfo() call to verify if bucket exists or not on
a regular basis. This has a large performance cost when there
are tons of servers involved.
We did optimize this part by vectorizing the bucket calls,
however its not enough, beyond 100 nodes and this becomes
fairly visible in terms of performance.
Removes the bloom filter since it has so limited usability, often gets saturated anyway and adds a bunch of complexity to the scanner.
Also removes a tiny bit of CPU by each write operation.
This commit replaces `ioutil.TempDir` with `t.TempDir` in tests. The
directory created by `t.TempDir` is automatically removed when the test
and all its subtests complete.
Prior to this commit, temporary directory created using `ioutil.TempDir`
needs to be removed manually by calling `os.RemoveAll`, which is omitted
in some tests. The error handling boilerplate e.g.
defer func() {
if err := os.RemoveAll(dir); err != nil {
t.Fatal(err)
}
}
is also tedious, but `t.TempDir` handles this for us nicely.
Reference: https://pkg.go.dev/testing#T.TempDir
Signed-off-by: Eng Zer Jun <engzerjun@gmail.com>
However, this slice is also used for closing the writers, so close is never called on these.
Furthermore when an error is returned from a write it is now reported to the reader.
bonus: remove unused heal param from `newBitrotWriter`.
* Remove copy, now that we don't mutate.
This PR adds deadlines per Write() calls, such
that slow drives are timed-out appropriately and
the overall responsiveness for Writes() is always
up to a predefined threshold providing applications
sustained latency even if one of the drives is slow
to respond.
major performance improvements in range GETs to avoid large
read amplification when ranges are tiny and random
```
-------------------
Operation: GET
Operations: 142014 -> 339421
Duration: 4m50s -> 4m56s
* Average: +139.41% (+1177.3 MiB/s) throughput, +139.11% (+658.4) obj/s
* Fastest: +125.24% (+1207.4 MiB/s) throughput, +132.32% (+612.9) obj/s
* 50% Median: +139.06% (+1175.7 MiB/s) throughput, +133.46% (+660.9) obj/s
* Slowest: +203.40% (+1267.9 MiB/s) throughput, +198.59% (+753.5) obj/s
```
TTFB from 10MiB BlockSize
```
* First Access TTFB: Avg: 81ms, Median: 61ms, Best: 20ms, Worst: 2.056s
```
TTFB from 1MiB BlockSize
```
* First Access TTFB: Avg: 22ms, Median: 21ms, Best: 8ms, Worst: 91ms
```
Full object reads however do see a slight change which won't be
noticeable in real world, so not doing any comparisons
TTFB still had improvements with full object reads with 1MiB
```
* First Access TTFB: Avg: 68ms, Median: 35ms, Best: 11ms, Worst: 1.16s
```
v/s
TTFB with 10MiB
```
* First Access TTFB: Avg: 388ms, Median: 98ms, Best: 20ms, Worst: 4.156s
```
This change should affect all new uploads, previous uploads should
continue to work with business as usual. But dramatic improvements can
be seen with these changes.
Design: https://gist.github.com/klauspost/025c09b48ed4a1293c917cecfabdf21c
Gist of improvements:
* Cross-server caching and listing will use the same data across servers and requests.
* Lists can be arbitrarily resumed at a constant speed.
* Metadata for all files scanned is stored for streaming retrieval.
* The existing bloom filters controlled by the crawler is used for validating caches.
* Concurrent requests for the same data (or parts of it) will not spawn additional walkers.
* Listing a subdirectory of an existing recursive cache will use the cache.
* All listing operations are fully streamable so the number of objects in a bucket no
longer dictates the amount of memory.
* Listings can be handled by any server within the cluster.
* Caches are cleaned up when out of date or superseded by a more recent one.
Add context to all (non-trivial) calls to the storage layer.
Contexts are propagated through the REST client.
- `context.TODO()` is left in place for the places where it needs to be added to the caller.
- `endWalkCh` could probably be removed from the walkers, but no changes so far.
The "dangerous" part is that now a caller disconnecting *will* propagate down, so a
"delete" operation will now be interrupted. In some cases we might want to disconnect
this functionality so the operation completes if it has started, leaving the system in a cleaner state.
- Implement a new xl.json 2.0.0 format to support,
this moves the entire marshaling logic to POSIX
layer, top layer always consumes a common FileInfo
construct which simplifies the metadata reads.
- Implement list object versions
- Migrate to siphash from crchash for new deployments
for object placements.
Fixes#2111
By monitoring PUT/DELETE and heal operations it is possible
to track changed paths and keep a bloom filter for this data.
This can help prioritize paths to scan. The bloom filter can identify
paths that have not changed, and the few collisions will only result
in a marginal extra workload. This can be implemented on either a
bucket+(1 prefix level) with reasonable performance.
The bloom filter is set to have a false positive rate at 1% at 1M
entries. A bloom table of this size is about ~2500 bytes when serialized.
To not force a full scan of all paths that have changed cycle bloom
filters would need to be kept, so we guarantee that dirty paths have
been scanned within cycle runs. Until cycle bloom filters have been
collected all paths are considered dirty.
When formatting a set validate if a host failure will likely lead to data loss.
While we don't know what config will be set in the future
evaluate to our best knowledge, assuming default settings.
This change provides new implementations of the XL backend operations:
- create file
- read file
- heal file
Further this change adds table based tests for all three operations.
This affects also the bitrot algorithm integration. Algorithms are now
integrated in an idiomatic way (like crypto.Hash).
Fixes#4696Fixes#4649Fixes#4359
* Prevent unnecessary verification of parity blocks while reading erasure
coded file.
* Update klauspost/reedsolomon and just only reconstruct data blocks while
reading (prevent unnecessary parity block reconstruction)
* Remove Verification of (all) reconstructed Data and Parity blocks since
in our case we are protected by bit rot protection. And even if the
verification would fail (essentially impossible) there is no way to
definitively say whether the data is still correct or not, so this call
make no sense for our use case.
xl.storageDisks is sometimes passed to some low-level XL functions. Some disks in
xl.storageDisks are set to nil when they encounter some errors. This means all
elements in xl.storageDisks will be nil after some time which lead to an unusable XL.
Currently `xl.json` saves algorithm information for bit-rot
verification. Since the bit-rot algo's can change in the
future make sure the erasureReadFile doesn't default to
a particular algo. Instead use the checkSumInfo.
* unit-tests: Unit tests for erasureCreateFile and erasureReadFile.
* appendFile() should return errXLWriteQuorum.
* TestErasureReadFileOffsetLength() tests erasureReadFile() for different offset and lengths.
* Fix for the failure seen in the erasure read unit test case. Issue #2227
* Move common erasure setup code to newErasureTestSetup()
* Review fixes. Add few more test cases for erasureReadFile.
Harshavardhana