v0.9.0-beta10

Changelog

381f778 Added possibility to create a number of indexes
36d1af9 Bump fluxdb for updated rules to index or not a tablet
12da9aa Fixed migrate command line tool not working anymore
65ac689 Fixed wrong resolver for TokenMeta TotalSupply, it was calling MaximumSupply
cb65466 Removed private dependency pulled in by mistake (go mod tidy does not work)
7ac62f2 Updated dgraphql to pull in streamingfast/dgraphql#2
dd4d6d7 Updated dgraphql, dgrpc and firehose
4c2d9f6 bump dstore to implement STORE_S3_READ_ATTEMPTS
be85fa6 bump kvdb library to raise default grpc recv max size (4MiB to 100MiB is enough for eosio chains)
2a801a7 bump req timeout for statedbproxy
7e8d4d3 bump search for longer transfer time on indexer upload (10m)
d9724ec regenerate a few riceboxes, and autogen files

v0.9.0-beta9

System Administration Changes

Added

• Added support for environment variable EOSWS_PUSH_V1_OUTPUT=true that forces push-transaction (guarantee:in-block) to output the same content format as nodeos 2.0.x (with Inlines)
• Added support for environment variable DSTORE_S3_BUFFERED_READ=true that forces reading S3 files (ex: blocks) ahead of processing, useful when S3 provider has trouble keeping long connections alive.
• Added support for looking up irreversible blocks on blockmeta (when the LIB was stuck for a while) from tokenmeta and trxdb-loader. They now use '--common-blockmeta-addr' flag if available
• Added --common-chain-core-symbol flag to define actual chain core symbol in the form <precision>,<symbol code> defaults to 4,EOS by default.
• Added --tokenmeta-readiness-max-latency with default=5m, now tokenmeta will show as "NotServing" through grpc healthcheck if last processed block (HEAD) is older than this. Value of 0 disables that feature.
• Added --relayer-source-request-burst with default=90 to allow a relayer connecting to another relayer to request a 'burst'
• Added --statedb-disable-indexing to disable indexing of tablet and injecting data into storage engine developer option, don't use that in production.
• Added --eosws-nodeos-rpc-push-extra-addresses to allow providing a list of backup EOS addresses when push-transaction does not succeed in getting the transaction inside a block (with push_guarantee)
• Added --eosws-max-stream-per-connection to allow changing how many stream can be active at the same time for a given WebSocket connection, defaults to 12 which was the hard-coded value.

• Added --eosws-statedb-proxy-retries, Number of time to retry proxying a request to statedb before failing (default 2)
• Added --eosws-nodeos-rpc-proxy-retries, Number of time to retry proxying a request to statedb before failing (default 2)

• Added --mindreader-max-console-length-in-bytes which is the limit in bytes that we allow action trace's console output to be before truncating them.
• Environment variable MINDREADER_MAX_TOKEN_SIZE can now be set to override bufio.Scanner() max token size (default 52428800, i.e. 50Mb) for EOSIO chains with huge transactions
• Flag --accounthist-mode to specific the accounthist mode of operation
• Added tools check accounthist-shards to
• Flag --common-include-filter-expr, --common-exclude-filter-expr, --common-system-actions-include-filter-expr can optionally specify multiple values, separated by ;;; and prefixed by #123; where 123 is a block number at which we stat applying that filter
• Added accounthist tools allows you to scan and read accounts dfuseeos tools accounthist read ... dfuseeos tools accounthist scan ...
• Flag --search-router-truncation-low-block-num to make the router aware of lower-block-truncation and serve requests accordingly
• Flag --mindreader-oneblock-suffix that mindreaders can each write their own file per block without competing for writes. #140
• Flag --eosws-disabled-messages a comma separated list of ws messages to disable.
• Flag --common-system-shutdown-signal-delay, a delay that will be applied between receiving SIGTERM signal and shutting down the apps. Health-check for eosws and dgraphql will respond 'not healthy' during that period.

Removed

• Breaking Change Removed --eosq-price-ticker-name flag, if you were using this flag, please use --common-chain-core-symbol instead to define it.
• Removed dgraphql-graceful-shutdown-delay, it was a left-over, unused. Must use --common-system-shutdown-signal-delay now
• Removed relayer-max-drift (now dependent on a new condition of presence of a "block hole", and no new block sent for 5 seconds)
• Removed relayer-init-time (no need for it with this new condition ^)

Changed

• The --eosq-available-networks logo field each network now has a maximum height of 70px.
• The --eosq-available-networks config of each network now accepts a logo_text that when present, is displayed alongside the logo field. This field is taken into consideration only when logo is defined. In this mode, the logo is fixed to 48px x 48px. If the logo_text value is eosq, this is rendered like the standard eosq logo.
• Breaking Change Changes to --eosq-available-networks config might be required around the logo field each network. You must now remove the logo field if it's not pointing to an existing image otherwise, the logo will not be rendered correctly.
• Applying a block filter over previously-filtered-blocks does not panic anymore, it applies the new filter on top of it, only if that specific filter has never been applied before. Applied filters definitions are concatenated in the block metadata, separated by ;;;
• Default trxdb-loader-batch-size changed to 100, Safe to do so because it does not batch when close to head.
• Improved relayer mechanics: replaced "max drift" detection by "block hole" detection and recovery action is now to restart the joining source (instead of shutting down the process)
• Improved dfuseeos tools check statedb-reproc-injector output by showing all shard statistics (and not just most highest block).
• Breaking Change Changed --statedb-enable-pipeline flag to --statedb-disable-pipeline to make it clearer that it should not be disable, if you were using the flag, change the name and invert the logical value (i.e. --state-enable-pipeline=false becomes --state-disable-pipeline=true)

Fixed

• Fixed validation of transaction ID passed to WebSocket get_transaction API, the prior validation was too permissive.
• Fixed a bug making search-forkresolver useless, because ignored by search-router.
• Fixed a bug on StateDB server not accepting symbol and symbol code as scope parameter value.
• Fixed shutdown on dgraphql (grpc/http) so it closes the active connections a little bit more gracefully.
• Fixed a bug in TiKV store implementation preventing it to delete keys correctly.
• Fixed a bug in eosws WebSocket get_transaction_lifecycle where a transaction not yet in the database would never stream back any message to the client.
• Fixed a bug with --mindreader-no-blocks-log option actually not being picked up (always false)
• Fixed a bug with /state/table/row not correctly reading row when it was in the table index.
• Fixed a bug with /state/tables/scopes where the actual block num used to query the data was incorrect leading to invalid response results.
• Fixed a bug with gRPC dfuse.eosio.statedb.v1/State#StreamMultiScopesTableRows where the actual block num used to query the data was incorrect leading to invalid response results.
• Fixed issue when reading ABI from StateDB where speculative writes were not handled correctly.
• Fixed issue when reading Table Row from StateDB where speculative writes were not handled correctly.
• Fixed a potential crash when reading ABI from StateDB and it does not exist in database.

v0.1.0-beta8

Changelog

4b40b01 bump fluxdb to fix netkv issue in statedb
437fe37 inform CHANGELOG

v0.1.0-beta7

Changelog

45706da fix netkv plugin missing registration

v0.1.0-beta6

This is a bugfix release (see v0.1.0-beta5 for complete changelog)

Bugfixes

d281ca0 Added missing updateauth and deleteauth action in system included actions

Doc improvements

a1e6f9c Improved README, merged some files together and improved filtering indexed actions
4310ef5 add filtering link to admin-guide
a1f4b15 adjust requirements check: yarn 1.10 is well enough
9e25253 bump generated bindata
26efd79 small tweaks to the init file output

v0.1.0-beta5

[v0.1.0-beta5] 2020-08-26

PUBLIC API Changes

Added

• Added tokens, accountBalances, tokenBalances calls to dgraphql (based on tokenmeta)

System Administration Changes

Changed

• Breaking Change FluxDB has been extracted to a dedicated library (github.com/dfuse-io/fluxdb) with complete re-architecture design.
• Breaking Change FluxDB has been renamed to StateDB and is incompatible with previous written data. See FluxDB Migration section below for more details on how to migrate.
• merger startblock behavior changed, now relies on state-file, see streamingfast/merger#1
• Changed merger-seen-blocks-file flag to merger-state-file to reflect this change.
• merger now properly handles storage backend errors when looking for where to start
• mindreader now automatically produces "merged blocks" instead of "one-block-files" when catching up (based on blocktime or if a blockmeta is reachable at --common-blockmeta-addr)
• mindreader now sets optimal EOS VM settings automatically if the platform supports it them when doing dfuseeos init.
• Changed --abicodec-export-cache-url flag to abicodec-export-abis-base-url and will contain only the URL of the where to export the ABIs in JSON.
• Changed --abicodec-export-cache flag to abicodec-export-abis-enabled.

Added

• Added tokenmeta app, with its flags
• Added support for filtered blocks, search indices and trxdb, with --common-include-filter-expr and --common-include-filter-expr.
• Added merged-filter app (not running by default), that generates filtered merged blocks files from regular merged blocks files.
• Added truncation handling to trxdb-loader, which will only keep a moving window of data in trxdb, and delete preceding transactions. Enabling that feature requires reprocessing trxdb.
  • --trxdb-loader-truncation-enabled
  • --trxdb-loader-truncation-window
  • --trxdb-loader-truncation-purge-interval
• Added --statedb-reproc-shard-scratch-directory to run StateDB reprocessing sharder using scratch directory to reduce RAM usage
• Added --merger-one-block-deletion-threads (default:10) to allow control over one-block-files deletion parallelism

• Added --merger-max-one-block-operations-batch-size to allow control over one-block-files batches that are looked up on storage,

• Added --eosws-with-completion (default: true) to allow control over that feature
• Added --mindreader-merge-threshold-block-age when processing blocks with a blocktime older than this threshold, they will be automatically merged: (default 12h)
• Added --mindreader-batch-mode to force always merging blocks (like --mindreader-merge-and-store-directly did) AND overwriting existing files in destination.
• Added --mindreader-wait-upload-complete-on-shutdown flag to control how mindreader waits on upload completion when shutting down (previously waited indefinitely)
• Added --search-live-hub-channel-size flag to specific the size of the search live hub channel capacity
• Added --search-live-preprocessor-concurrent-threads: number of thread used to run file source preprocessor function
• Added --abicodec-export-abis-file-name, contains the URL where to export the ABIs in JSON
• Added --metrics-listen-addr to control on which address to server the metrics API (Prometheus), setting this value to an empty string disable metrics serving.
• Added --dashboard-metrics-api-addr to specify a different API address where to retrieve metrics for the dashboard.
• Added Experimental support for kvdb backend netkv://, an extremely simple network layer over badger to allow running dfuse components in separate instances.

Removed

• The --merger-delete-blocks-before flag is now removed and is the only behavior for merger.
• The --mindreader-merge-and-store-directly flag was removed. That behavior is now activated by default when encountering 'old blocks'. Also see new flag mindreader-batch-mode.
• The --mindreader-discard-after-stop-num flag was removed, its implementation was too complex and it had no case where it was really useful.
• The --mindreader-producer-hostname flag was removed, this option made no sense in the context of mindreader app.
• The --eosq-disable-tokenmeta flag was removed, token meta is now included, so this flag is now obsolete.
• The --eosq-on-demand flag was removed, this was unused in the codebase.

Fixed

• Fixed issue where blockmeta was not serving on GRPC at all because it couldn't figure out where to start on the stream
• Fixed issue with merger with a possible panic when reading a one-block-file that is empty, for example on a non-atomic storage backend
• Fixed issue with mindreader not stopping correctly (and showing any error) if the bootstrap phase (ex: restore-from-snapshot) failed.
• Fixed issue with pitreos not taking a backup at all when sparse-file extents checks failed.
• Fixed issue with dfuseeos tools check merged-blocks (start/end block, false valid ranges when the first segment is not 0, etc.)
• Improved performance by using value for bstream.BlockRef instead of pointers and ensuring we use the cached version.
• mindreader and node-manager improved nodeos log handling

FluxDB to StateDB Migration

FluxDB required an architecture re-design to fit with our vision about the tool and make it chain agnostic (so
it is easier to re-use on our other supported chain).

The code that was previously found here has been extracted to its own library (https://github.com/dfuse-io/fluxdb).
There is now a new app named StateDB (statedb is the app identifier) in dfuse for EOSIO that uses FluxDB to
support all previous API endpoints served by the FluxDB app as well as now offering a gRPC interface.

While doing this, we had to change how keys and data were written to the underlying engine. This means that all
your previous data stored cannot be read anymore by the new StateDB and that new data written by StateDB will
not be compatible on a previous instance.

What that means exactly is that StateDB will require to re-index all the current merged blocks up to live blocks
before being able to serve requests. This is the main reason why we decided to rename the app, so you are forced
to peform this step.

Here the steps required to migrate to the new statedb app:

• In your dfuse.yaml config, under replace the fluxdb app by statedb and all flags prefixed with
  fluxdb- must now be prefixed with statedb-.

  From:

  start:
  args:
  - ...
  - fluxdb
  - ...
  flags:
    ...
    fluxdb-http-listen-addr: :9090
    ...

  To:

  start:
  args:
  - ...
  - statedb
  - ...
  flags:
    ...
    statedb-http-listen-addr: :9090
    ...

• If you had a customization for fluxdb-statedb-dsn, you must first renamed it statedb-store-dsn
  to statedb-store-dsn. Important You must use a new fresh database, so update your argument
  so it points to a new database, ensuring we don't overwrite data over old now incompatible data.
  If you did not customize the flag, continue reading, the default value has changed to point to a
  fresh storage folder.

  If you had a customization for eosws-flux-addr, rename to eosws-statedb-grpc-addr and ensure it
  points to the StateDB GRPC address (and not the HTTP address), its value must be the same as the flag
  statedb-grpc-listen-addr.

• If you have a custom fluxdb-max-threads flag, removed it, customizing this value is not supported
  anymore.

• From an operator standpoint, what we suggest is to craft a dfuse.yaml config that starts StateDB
  only in inject mode only. You let this instance run until StateDB reaches the live block of your
  network.

  Once you have reached this point, you can now perform a switch to the new StateDB database. Stop
  the injecting instance. Stop your production instance, renaming old app id fluxdb to statedb
  (and all flags) then reconfigure it so the statedb-store-dsn points to the database populated
  by the injecting instance. At this point, you can restart your production node and continue
  normally using the new statedb app.

v0.1.0-beta4

RELEASE DATE: 2020-06-23

We’re constantly working on improving the user experience with each release, so here’s what changed since our last update:

Public API Changes

On the Public API side of things, which is meant for developers querying the data on chain through the dfuse hosted services or through a local version of the dfuseeos product, we introduced 1 BREAKING change for 2 fields;

Changed

• BREAKING: For more clarity when using these fields, the eosws transaction lifecycle fields creation_irreversible & cancelation_irreversible were renamed respectively to dtrx_creation_irreversible and dtrx_cancelation_irreversible.

System Administration Changes

On the side of System Administration, which is meant for developers running a local version of the dfuseeos product, we introduced many changes, some of them BREAKING;

Added

• App: mindreader-stdin — This app relays blocks and produces one-block-files (or merged-blocks-files) based on stdin, without trying to manage nodeos. This is an alternative way of seeding your dfuse system if you have existing tooling for managing nodeos operations. It uses only a subset of the mindreader flags and does not stop on TERM signal until it receives an EOF signal from stdin.
• App: booter — We’re especially excited about this new app! This app allows a boot sequence to be applied to a new chain (see the README for more information). It also brings new flags with it:
  • --booter-bootseq — File path to the desired boot sequence (default: "./bootseq.yaml").
  • --booter-data-dir — Booter's working directory (default: "{dfuse-data-dir}/booter").
  • --booter-nodeos-api — Target API address to communicate with underlying nodeos (default: “http://localhost:8888/”).
  • --booter-private-key — Genesis private key having control over an eosio account to be able to create all related system accounts & contracts.
  • --booter-vault-file — Wallet file that contains encrypted key material.
• Command: dmesh — This command was added to the dfuseeos tools with 2 new flags: --dsn & --service-version. This command allows you to inspect dmesh search peers. It currently only supports etcd (the server in which all the information is stored, see this graphic).
• Command: kv — This command was added to the dfuseeos tools with 8 new subcommands: get, scan, prefix, account, blk, blkirr, trx, trxtrace. This command allows you to retrieve data from trxdb at a lower lever (e.g. only rows of data).
• Command: db — This command was added to the dfuseeos tools with 2 new subcommands: blk, trx. This command allows you to retrieve data from trxdb at a higher level (e.g. assembling multiple rows into a data structure).
• Command: check trxdb-blocks — This command was added to the dfuseeos tools and ensures linearity of irreversible blocks in storage, which is useful to know if you've missed some block ranges when doing parallel insertions into your trxdb storage. This relates to trxdb-loader's job being done right.
• Flag: --mindreader-fail-on-non-contiguous-block — This flag lets you enable the mindreader ContinuityChecker (default:false). See mindreader under the CHANGED section below.
• Flag: --log-level-switcher-listen-addr — This flag lets you change the port to which you send your HTTP request to switch the log levels of a given component in real-time (default: ":1065") (see LOGGING.md).
• Flag: --common-ratelimiter-plugin — This flag lets you enable a rate limiter plugin (default: "null://").
• Flag: --pprof-listen-addr — This flag enables you to change the port listening to pprof (default: ":6060").
• Flag: --search-common-dfuse-events-unrestricted — This flag lets you lift all restrictions on Indexing Limits for searching dfuse Events (max field count, max key length, max value length).
• Flag: --mindreader-discard-after-stop-num — This flag lets you discard all blocks after stop-num if it is set to true (default:false). See the mindreader bullet below.
• mindreader now writes remaining one-block files after stop-block if --mindreader-merge-and-store-directly is set, unless --mindreader-discard-after-stop-num is set to true. This improves the experience of a user following PARTIAL_SYNC.md steps, producing merged files up to a certain block, then switching to one-block files with a separate merge instance. This improves the parallel reprocessing experience.
• The auto-backups feature will now only trigger if os.Hostname() returns the non-empty value of one of these new flags:
  • --node-manager-auto-backup-hostname-match
  • --node-manager-auto-snapshot-hostname-match
  • --mindreader-auto-backup-hostname-match
  • --mindreader-auto-snapshot-hostname-match
• The trxdb “deduper” now reduces storage by removing identical action data and calls the "reduper" to add this data back. EOS Mainnet saves 20%(!) off of storage because of this.

Changed

• BREAKING: eosdb was renamed to trxdb (the storage location where blocks & transactions are saved). Verify and adjust your command arguments and 'dfuse.yaml' file for references to eosdb. Specifically, eosdb’s healthz endpoint field is now trxdb’s healthz endpoint field, so you might need to adjust your monitoring.
• BREAKING: The search flag --search-common-dfuse-hooks-action-name was renamed to --search-common-dfuse-events-action-name.
• BREAKING: The abicodec app default value for its --abicodec-cache-base-url and --abicodec-export-cache-url flags was changed to {dfuse-data-dir}/storage/abicache (fixing a typo in abicahe). To remain compatible, simply do a manual rename on disk before starting the updated version such as mv {dfuse-data-dir}/storage/abicahe {dfuse-data-dir}/storage/abicache.
• BREAKING: The fluxdb flag --fluxdb-enable-dev-mode was removed. Use --fluxdb-enable-live-pipeline=false to get the same behavior as before.
• mindreader ContinuityChecker is not enabled by default anymore.
• The value of node-manager and mindreader’s respective --shutdown-delay flags is now also applied to commands like snapshot or backup, so they become "not-ready" on /healthz endpoint, allowing a load-balancer to take it out of the pool before they actually stop working.
• The dfuseeos tools check blocks command was renamed to dfuseeos tools check merged-blocks.
• The search roarCache is now based on a normalized version of the query string (e.g. a:foo b:bar is now equivalent to b:bar a:foo, etc.). Note that this will make previously-cached entries useless.
• The --node-manager-number-of-snapshots-to-keep and --mindreader-number-of-snapshots-to-keep flags now default to 0, which means an unlimited number of snapshots will be kept, instead of their previous default of 5 which only kept the latest 5 backups.
• Various startup speed improvements for the blockmeta, bstream, and search-indexer apps.

Removed

• The --search-indexer-num-blocks-before-start flag was removed from search-indexer. Because of that, search-indexer now automatically resolves its start block.

Fixed

• trxdb now correctly implements "BatchGet" on most operations, giving a great(!) performance increase over previous versions.
• Global flags and specifically the --search-common-dmesh-dsn flag are now correctly parsed from the config file.
• search-indexer no longer overflows on negative startblocks on new chains, it fails fast instead.
• search-archive relative-start-block truncation now works.
• search-forkresolver no longer throws a nil pointer (app was previously broken).
• mindreader now has a more resilient shutdown handling (expects EOF on nodeos stdout).
• The mindreader and node-manager logs from nodeos that go through zap now have their level parsed correctly instead of all being seen as DEBUG.

v0.1.0-beta3

System Administration Changes

Breaking changes

• To improve dfuse instrumented nodeos binary processing speed, we had to make incompatible changes to data exchange format going out of nodeos. This requires you to upgrade your dfuse instrumented nodeos binary to latest version (https://github.com/dfuse-io/eos/releases/tag/v2.0.5-dm-12.0). Follow instructions in at https://github.com/dfuse-io/dfuse-eosio/blob/develop/DEPENDENCIES.md#dfuse-instrumented-eosio-prebuilt-binaries to install the latest version for your platform.
• --mindreader-working-dir default value is now {dfuse-data-dir}/mindreader/work instead of {dfuse-data-dir}/mindreader this is to prevent mindreader from walking files into the working dir and trying to upload and delete nodes system files like fork_db.dat
• The Changed section below lists a bunch of breaking changes regarding command-line flags.

Added

• Added --eosq-available-networks json string to configure the network section of eosq.

     {
       "id": "id.1",
       "name": "Network Name",
       "is_test": false,
       "logo": "/images/network-logo.png",
       "url": "https://www.example.com/"
     },
   ]

• Added --eosq-default-network string to configure the default network eosq
• Added --eosq-disable-analytics bool to configure eosq analytics
• Added --eosq-display-price bool to configure if eosq displays prices
• Added --eosq-price-ticker string to configure if eosq price ticker
• Added --eosq-on-demand bool to configure if eosq serves an on-demand network
• Added --eosq-disable-tokenmeta bool to configure if eosq disables tokenmenta
• Added --eosq-environment environment where eosq will run (local, dev, production)
• Added --apiproxy-autocert-domains, --apiproxy-autocert-cache-dir and --apiproxy-https-listen-addr to serve SSL directly from proxy.

Removed

• Removed the --merger-store-timeout flag. Not needed anymore, as some sensible timeouts have been put here and there, using the latest [email protected] that is context-aware.

Changed

• We improve by 4x times the performance of dfuse instrumented nodeos binary on heavy EOS Mainnet blocks. This required changes to nodeos data exchange format, so you will need to upgrade it, see the Breaking change entry at top of this section.
• Fixed behavior of --eosq-api-endpoint-url to allow specifying protocol (ex: https://api.mydomain.com)
• The kvdb-loader application was renamed trxdb-loader. In general what was (confusingly) named kvdb is now trxdb, so that kvdb can now take on its full meaning of a lean key-value storage abstraction (which is also used by FluxDB).
  • All --kvdb-loader flags have been renamed to --trxdb-loader.
  • Metrics ID for kvdb-loader has been changed to trxdb-loader (check your dashboards)
• The --mindreader-merge-and-upload-directly was renamed to --mindreader-merge-and-store-directly.
• --common-blocks-store-url now replaces all of these flags:
  • --mindreader-merged-blocks-store-url, --relayer-blocks-store, --fluxdb-blocks-store, --kvdb-loader-blocks-store, --blockmeta-blocks-store, --search-indexer-blocks-store, --search-live-blocks-store, --search-forkresolver-blocks-store, --eosws-blocks-store
• --common-oneblock-store-url now replaces these flags:
  • --mindreader-oneblock-store-url, --merger-one-block-path
• --common-backup-store-url now replaces these flags:
  • --node-manager-backup-store-url, --mindreader-backup-store-url
• --search-common-indices-store-url now replaces these flags:
  • --search-indexer-indices-store, --search-archive-indices-store
• --common-blockstream-addr now replaces these flags:
  • --fluxdb-block-stream-addr, --kvdb-loader-block-stream-addr, --blockmeta-block-stream-addr, --search-indexer-block-stream-addr, --search-live-block-stream-addr, --eosws-block-stream-addr
• --common-blockmeta-addr now replaces these flags:
  • --search-indexer-blockmeta-addr, --search-router-blockmeta-addr, --search-live-blockmeta-addr, --eosws-block-meta-addr, --dgraphql-block-meta-addr
• --common-network-id now replaces this flag: --dgraphql-network-id
• --common-auth-plugin now replaces these flags:
  • --dgraphql-auth-plugin, --eosws-auth-plugin
• --fluxdb-statedb-dsn replaces --fluxdb-kvdb-store-dsn (to avoid confusion between what's actually stored in kvdb and how FluxDB is using it (as a simple kv store).
• --trxdb-loader-parallel-file-download-count replaces --kvdb-parallel-file-download-count
• --common-trxdb-dsn replaces these flags:
  • --blockmeta-kvdb-dsn, --abicodec-kvdb-dsn, --eosws-kvdb-dsn, --dgraphql-kvdb-dsn, --kvdb-loader-kvdb-dsn
• Changed default value for storage URL for fluxdb and kvbd (now named trxdb)

Fixed

• --kvdb-loader-chain-id not being taken into account. This affected the decoding of public keys during the kvdb loading process.

v0.1.0-beta2

Added

• Added apiproxy application, with its flags
• Added --log-format option for JSON output and log-to-file bool (default to true, same behavior as before)
• Filtering (whitelist and blacklist) of what is indexed in Search, based on Google's Common Expression Language. See details here. Added --search-common-action-filter-on-expr and --search-common-action-filter-out-expr.
  • NOTE: This doesn't affect what is extracted from the chain, allowing you to re-index selectively without a chain replay.

Changed

• CLI: dfuseeos init now writes dfuse.yaml with the start command's flags, also the array of components to start
• CLI: new {dfuse-data-dir} replacement string in config flags, also changed default flag values
• --node-manager-config-dir now ./producer (was manager/config)
• --node-manager-data-dir now {dfuse-data-dir}/node-manager/data (was managernode/data)
• --mindreader-config-dir now ./mindreader (was mindreadernode/config)
• --mindreader-data-dir now {dfuse-data-dir}/mindreader/data (was mindreadernode/data)
• CLI: regrouped some flags:
• --search-indexer-dfuse-hooks-action-name, --search-live-dfuse-hooks-action-name, --search-forkresolver-dfuse-hooks-action-name fused into new --search-common-dfuse-hooks-action-name.
• --search-...-mesh-publish-polling-duration fused into new --search-common-mesh-publish-polling-duration.
• all of the --search-mesh-... options were renamed to --search-common-mesh-... (previously --search-mesh-service-version, --search-mesh-namespace, --search-mesh-store-addr)
• dashboard: now separate metrics for mindreader vs producer node
• dashboard doesn't act as a reverse proxy anymore (apiproxy does)
• dashboard's default port is now :8081
• eosq's port is now proxied through :8080, so use that.
• App manager renamed to node-manager. All of its flags were changed from --manager-... to --node-manager-...

Removed

• The --search-...-indexing-restrictions-json. This was replaced by the filtering listed above.

v0.1.0-beta1

Initial release