* Database Versioning: The behavior is to load the database at the existing version if it
already exists or create the newest version if it doesn't
* multi-column by default: This makes the separated world state storage column required by mark sweep on by default
Signed-off-by: Adrian Sutton <adrian.sutton@consensys.net>
Windows has some slightly different Java NIO semantics, regarding size
after writing and whether files are deleted on close (they aren not).
The first issue is we shouldn't be using Integer.SIZE when we mean
Integer.BYTES.
The second issue is we cannot count on these work files showing up or
being deleted from the file system consistently across platforms. The
ordering is consistent within platforms but not across. The test was
re-written to check the read and write file numbers instead.
Signed-off-by: Adrian Sutton <adrian.sutton@consensys.net>
* Mark Sweep Pruner
* add `unload` method on Node interface, which is a noop everywhere but on `StoredNode`
Signed-off-by: Adrian Sutton <adrian.sutton@consensys.net>
* [PAN-2853] Create a metric tracking DB size
- create a Prometheus long gauge to track the database size
- `rocksdb.live-sst-files-size` : 872a261ffc/include/rocksdb/db.h (L685)
- use `rocksdb.live-sst-files-size` property rather than `rocksdb.total-sst-files-size` (performance purpose: see warning 872a261ffc/include/rocksdb/db.h (L680))
- add tests to check creation of metrics
* fix wildcard import
Signed-off-by: Adrian Sutton <adrian.sutton@consensys.net>
* Columnated storage to allow for iteration over world state
* change MetricsCategory to PantheonMetricsCategory
* consistency renaming of kvstores
Signed-off-by: Adrian Sutton <adrian.sutton@consensys.net>
Replaces the RocksDB based queue for pending world state download tasks with one that uses a simple file. Added tasks are appended to the file while the reader starts from the beginning of the file and reads forwards.
Periodically a new file is started to limit the disk space used. The reader deletes files it has completed reading.
Signed-off-by: Adrian Sutton <adrian.sutton@consensys.net>
Default values are currently unchanged but increasing both these values to 4 appears to improve sync performance for both ropsten and Mainnet fast syncs.
Signed-off-by: Adrian Sutton <adrian.sutton@consensys.net>
Introduce a pipeline based full sync process. Currently toggled off but can be enabled via a --X option.
Signed-off-by: Adrian Sutton <adrian.sutton@consensys.net>
Move RocksDBStats to it's own module. This also brings metrics to
metrics:core since none of our other module have nested modules but
they have peer modules.
Signed-off-by: Adrian Sutton <adrian.sutton@consensys.net>
* Use labelled timer to differentiate between metrics of public and private db
* Update the label names
Signed-off-by: Adrian Sutton <adrian.sutton@consensys.net>
Adds handling for options that need to be configurable for tuning and performance but in general are not long-term supported configuration options.
Options that are added via a mixin, are hidden, and that start with `--X` are considered 'Unstable Options.' These options do not show up via normal help facilities but are enumerated via a `XHelp` CLI invocation. These options are unstable, unsupported, and undocumented. They may be added, removed, or modified without prior warning or notice. Use them at your own risk.
For example, the `--Xrocksdb-max-open-files` flag can be set via CLI and can increase the number of max open files RocsDB uses. On a 8GB memory system the default is appropriate, but on a 128GB memory system a much larger number may improve performance. Normal operation scenarios will not need to modify this flag.
Signed-off-by: Adrian Sutton <adrian.sutton@consensys.net>
Allows monitoring to accurately calculate the number of items in a pipe even after it has aborted.
Signed-off-by: Adrian Sutton <adrian.sutton@consensys.net>
* Number of metrics labels need to match up with constructor
* Number of labels must be consistant, so I split it into two metrics
* Also, naming best practices say that sum() and avg() of a metric
should be meaningful, separating into two metrics fixes that.
* fix style issues (finals, intellij warnings)
* Change NoOpMetrics to check label count.
* Cascading changes to support this in many support classes. Mostly places
we presumed all NoOpMetrics were equals.
Signed-off-by: Adrian Sutton <adrian.sutton@consensys.net>
Biggest change is that UnusedVariable and UnusedMethod went to WARN by
default. Since our build is a no warning build this means we either need
to turn them off or fix them. I mostly opted for the latter. Test code
was mostly fixed, unused loggers were deleted, and other shipped code
was mostly suppressed.
Two less noisy fixes to not use `SortedSet` and to use zero based
comparable results instead of -1, 0, and 1. Also a compiler nit in
errorprone was suppressed, per the description it won't affect us.
Signed-off-by: Adrian Sutton <adrian.sutton@consensys.net>
* Capture input and output metrics for each pipe rather than out input.
* Replace the batching processor with a BatchingReadPipe that wraps a normal pipe and creates batching on the fly.
Avoids needing an extra thread and synchronization overhead to create batches.
Signed-off-by: Adrian Sutton <adrian.sutton@consensys.net>
* Use a single thread to persist received world state nodes to avoid RocksDB timeouts due to contention on the write lock.
* Fix RocksDbTaskQueue so it doesn't read stale data when resuming a transfer.
* Skip downloading empty trie nodes.
* Log unhandled errors from world state download requests
Signed-off-by: Adrian Sutton <adrian.sutton@consensys.net>
* RocksDB uses synchronized methods so can use plain long instead of AtomicLong.
* Synchronize close method and use simple boolean instead of AtomicBoolean.
Signed-off-by: Adrian Sutton <adrian.sutton@consensys.net>
Gives approximately a 10% perf improvement in isolation and in real-world tests allows the world state requests to actually reach the concurrent limit.
Signed-off-by: Adrian Sutton <adrian.sutton@consensys.net>
* Skip requesting data we already have but continue to walk the tree to ensure we have all child nodes.
* Don't delete fast sync state on stop. Allow resuming world state downloads.
Signed-off-by: Adrian Sutton <adrian.sutton@consensys.net>
Two changes:
* Stream chains now take up less vertical lines, only breaking on
stream operations.
* Long annotations that span multiple lines no longer have a dangling
parentesis and indent 4 spaces.
Signed-off-by: Adrian Sutton <adrian.sutton@consensys.net>
Ratan Rai Sur
Danno Ferrin
mbaxter
Karim T
CJ Hare
Abdelhamid Bakhta
Adrian Sutton
Adrian Sutton
Danno Ferrin
Puneetha Karamsetty
Ivaylo Kirilov