This improves the performance of gather writes (p and q blobs) in
particular on high-latency links by reducing round-trips.
Tested by uploading Kopia working directory
(3 GB, files:32157 dirs:6051) to SFTP server on LAN (TrueNAS) over WiFi:
Before: 2m4s (~24 MB/s)
After: 1m13s (~41 MB/s)
From https://github.com/google/gvisor/tree/master/tools/checklocks
This will perform static verification that we're using
`sync.Mutex`, `sync.RWMutex` and `atomic` correctly to guard access
to certain fields.
This was mostly just a matter of adding annotations to indicate which
fields are guarded by which mutex.
In a handful of places the code had to be refactored to allow static
analyzer to do its job better or to not be confused by some
constructs.
In one place this actually uncovered a bug where a function was not
releasing a lock properly in an error case.
The check is part of `make lint` but can also be invoked by
`make check-locks`.
* feat: persisting retention options in repository blob
- plumb retention parameters through wrapped storage
- generalize aes encryption mechanism
- rewrite the retention blob on password change
- do not write retention blob when empty
* handle retention-blob not-found failures
* cli params to set retention modes on repository create
* enable versioned map mock storage with retention settings
* adding unit tests
* write format and retention blob with retention settings if available
* rename certain functions and constants specific to format blob
* delete retention cache on password-change
* fix: replace SetTime() api call with TouchBlob()
* Update repo/repository_test.go
Co-authored-by: Nick <nick@kasten.io>
* pr feedback and codecov improvements
* fix: rename retention-blob structures to generic blob-cfg
* fix: remove minio dependency on retention constants
Co-authored-by: Shikhar Mall <shikhar@kasten.io>
Co-authored-by: Nick <nick@kasten.io>
* upload: do not launch error group when doing non-parallel upload
* encryption: refactored code to reduce slice allocation, this uncovered a minor subtle bug in index recovery due to manual memory management
* content: fixed use-after-free bug during index recovery that started failing test after encryption memory management improvements
* gather: tweaked retention and chunk sizes
* gather: added typicalContiguousAllocator with 8MB chunks for default configuration
* content: fixed data race in IterateUnreferencedBlobs
* upload: fixed data race between uploader and estimator
* testing: fixed data race in repo/blob/logging test
* makefile: run tests on CI/linux/amd64 with -race
* robustness: fixed test race
* content: fixed data race getContentDataUnlocked that triggers TestParallelWrites - looks scary but in practice very hard to trigger in real life and does not cause data corruption
* testing: reduce test complexity under race detector
* server: fixed minor race in refreshStatus()
* testing: reduced depth of sharedTestDataDir2
* ci: run race detector in separate job
* ci: run unit test race detector in parallel to integration tests
, where blob.Storage.PutBlob gets a list of slices and writes them sequentially
* performance: added gather.Bytes and gather.WriteBuffer
They are similar to bytes.Buffer but instead of managing a single
byte slice, they maintain a list of slices that and when they run out of
space they allocate new fixed-size slice from a free list.
This helps keep memory allocations completely under control regardless
of the size of data written.
* switch from byte slices and bytes.Buffer to gather.Bytes.
This is mostly mechanical, the only cases where it's not involve blob
storage providers, where we leverage the fact that we don't need to
ever concatenate the slices into one and instead we can do gather
writes.
* PR feedback
