v0.39.2

NeoFS chain's auto-deployment fixes with some maintainability improvements.

Added

• Logs on a connection drop in the cache of NeoFS API clients (#2694)

Fixed

• Auto-deployment of the Balance and Container contracts (#2695)

v0.39.1

Fixed

• Fund transfer deadlock in NeoFS chain auto-deploy/update procedure (#2681)
• Invalid contracts' update transactions when epochs are stuck during the NeoFS chain update (#2680)
• Metrics availability during startup (#2677)

v0.39.0 - Baegado

Complete contract autodeploy/autoupdate functionality, much simplified SN/IR attribute interaction, numerous optimizations and obligatory bug fixes --- that's your new NeoFS release in short. Beware of deprecated commands and options removal, check your scripts and configurations and update to the latest and greatest of NeoFS.

Added

• Policer's setting to the SN's application configuration (#2600)
• Support of verified domains for the storage nodes (#2280)
• neofs-lens storage status CLI command (#2550)
• Human-readable output of objects' creation timestamp to neofs-cli container list-objects (#2653)
• Ability to preset P2PNotary and NeoFSAlphabet roles to validators at the FS chain's genesis (#2643)

Fixed

• neofs-cli netmap netinfo documentation (#2555)
• GETRANGEHASH to a node without an object produced GETRANGE or GET requests (#2541, #2598)
• Iteration over locally collected container estimation values (#2597)
• Synchronous objects PUT to the local storage (#2607)
• copies_number field was not used in PUT request handling (#2607)
• Neo-go notary deposit data workaround (#2429)
• Error of missing network endpoint in neofs-cli object commands when it is set in the file config (#2608)

Changed

• FSTree storage now uses more efficient and safe temporary files under Linux (#2566, #2624, #2633)
• BoltDB open timeout increased from 100ms to 1s (#2499)
• Internal container cache size from 10 to 1000 (#2600)
• Transaction witness scope no longer uses CustomGroups relying on more strict Rules now (#2619)
• New optimized SDK version is integrated (#2622)
• IR uses internal LOCODE DB from Go package (#2610, #2658)
• Contract group for the committee is no longer registered/used in the Sidechain auto-deployment (#2642)
• IR does not change SNs' attributes, SNs expand UN/LOCODE attributes do it themselves (#2612)
• The priority of metrics service is increased (#2428)
• The priority of running control service is maximized (#2585)

Removed

• Deprecated no-precheck flag of neofs-cli container set-eacl (#2496)
• Recently-handled objects Policer's cache (#2600)
• Deprecated blobovnicza substorage (#2614)
• Contract group wallet support from neofs-adm (#2619)
• neofs-cli util locode generateand neofs-cli util locode info commands (#2610)
• Locode DB configuration options (#2610)
• v prefix in version (#2640)

Updated

• Update minimal supported Go version up to v1.19 (#2485)
• Update neo-go to v0.104.0 (#2221, #2309, #2596, #2626, #2639, #2659)
• neofs-lens inspect object commands to get with inspect deprecation (#2548)
• Update tzhash to v1.7.1
• github.com/panjf2000/ants/v2 to v2.8.2
• github.com/nspcc-dev/neofs-contract to v0.18.0 (#2580)
• golang.org/x/net to 0.17.0 (#2621)
• github.com/nats-io/nats-server/v2 to 2.9.23 (#2623)
• UUID, golang-lru, compress, go-multiaddr, nats.go, cobra, grpc and other dependencies (#2632)
• hrw library to v2.0.0 version (#2629, #2658)
• github.com/nats-io/nkeys to 0.4.6 (#2636)
• neofs-contract to v0.19.1 (#2659)

Updating from v0.38.1

Blobovniczas are gone from the node with this release, see 0.38.0 upgrade notes for migration instruction (the tool is still provided, but will be gone after a couple of minor releases).
neofs-adm no longer creates a contract group wallet and no longer needs it for the operation which may affect deployment scripts.
Pre-0.39.0 storage nodes may not be able to register on the network running 0.39.0 inner ring nodes unless they provide complete LOCODE data in the configuration (0.39.0 SNs add it automatically, so upgrade SNs).
native_activations subsection of consensus IR configuration should be removed. It never actually did anything useful anyway.
Automatic contract deployments and updates require fschain_autodeploy configuration, currently it's optional an IR behaves as previously by default.

v0.38.1

A tiny fix for peapod-enabled nodes.

Fixed

• Inability to start node with peapods configured (#2576)
• neofs-adm morph set-config command error messages (#2556)

v0.38.0 - Gogado

A number of important fixes are brought by this release, including protocol ones (subnet and system EACL changes), as well as an updated storage subsystem and useful CLI updates making it more powerful and user-friendly at the same time.

Some previously deprecated configuration options were removed, so be careful and this is the last minor release to support the Blobovnicza tree storage subsystem. Starting with the next minor release, the component will be purged,so be prepared (see Updating section) and migrate to more efficient and simple Peapod.

Added

• Embedded Neo contracts in contracts dir (#2391)
• dump-names command for adm
• renew-domain command for adm
• Stored payload metric per container (#2116)
• Stored payload metric per shard (#2023)
• Histogram metrics for RPC and engine operations (#2351)
• New storage component for small objects named Peapod (#2453)
• New blobovnicza-to-peapod tool providing blobovnicza-to-peapod data migration (#2453)
• SN's version and capacity is announced via the attributes automatically but can be overwritten explicitly (#2455, #602)
• peapod command for neofs-lens (#2507)
• New CLI exit code for awaiting timeout (#2380)
• New CLI exit code for already removed objects (#2376)
• Validation of excessive positional arguments to neofs-cli commands (#1941)
• --lifetime flag to bearer create and object put CLI commands (#1574)
• --expired-at flag to session create and storagegroup put CLI commands (#1574)
• Sessions to RPC server running in IR's local consensus mode (#2532)
• neofs-cli object nodes command to get SNs for an object (#2512)
• Fetching container estimations via iterators to prevent NeoVM stack overflow (#2173)
• neofs-adm morph netmap-candidates CLI command (#1889)
• SN network validation (is available by its announced addresses) on bootstrap by the IR (#2475)
• Display of container alias fee info in neofs-cli netmap netinfo (#2553)
• neofs-lens storage inspect CLI command (#1336)
• neofs-lens payload-only flag (#2543)
• neofs-lens meta put CLI command (#1816)

Fixed

• neo-go RPC connection loss handling (#1337)
• Concurrent morph cache misses (#1248)
• Double voting for validators on IR startup (#2365)
• Skip unexpected notary events on notary request parsing step (#2315)
• Session inactivity on object PUT request relay (#2460)
• Missing connection retries on IR node startup when the first configured mainnet RPC node is not in sync (#2474)
• Storage node no longer ignores unhealthy shards on startup (#2464)
• Processing of status errors returned by API client from NeoFS SDK RC-9 (#2465)
• neofs-lens write-cache list command duplication (#2505)
• neofs-adm works with contract wallet in init and update-contracts commands only (#2134)
• Missing removed but locked objects in SEARCH's results (#2526)
• LOCK objects and regular objects expiration conflicts (#2392)
• SN responds with a different node information compared to a bootstrapping contract call's argument (#2568)
• neofs-cli object put command processes multiple --attributes flags (#2427)

Removed

• Deprecated morph.rpc_endpoint SN and morph.endpoint.client IR config sections (#2400)
• neofs_node_object_epoch metric for IR and SN (#2347)
• Subnets support (#2411)
• Logging utility completely replaced with zap.Logger (#696)
• System eACL modification ability in the neofs-cli and IR (netmap candidate validation) (#2531)

Changed

• CLI --timeout flag configures whole execution timeout from now (#2124)
• CLI default timeout for commands with --await flag increased to 1m (#2124)
• BlobStor tries to store object in any sub-storage with free space (#2450)
• SN does not store container estimations in-mem forever (#2472)
• CLI neofs-cli container set-eacl checks container's ownership (#2436)

Updated

• neofs-sdk-go to v1.0.0-rc.11

Updating from v0.37.0

CLI command timeouts (flag --timeout) now limit the total command execution time and not single network operation. If any commands suddenly start to fail on timeout, try increasing the value, for example, twice. Also note that the execution of commands with the --await flag and without an explicitly specified time period is now limited to 1 minute. This value can be changed with--timeout flag.

Histogram (not counter) RPC/engine operation handling time metrics were added. For an old engine *operation_name*_duration a new *operation_name*_time is available.For an old *operation_name*_req_duration RPC a new rpc_*operation_name*_time is available. The old ones (the counters) have been deprecated and will be removed with the following minor release.

Container estimation main node calculation has been changed. A new 32-byte long array is taken as a sorting pivot: it is an estimated container with the first 8 bytes replaced with a target epoch in a little-endian encoding now.

Deprecated morph.rpc_endpoint SN and morph.endpoint.client IR configurations have been removed. Use morph.endpoints for both instead.

Deprecated neofs_node_object_epoch metric for IR and SN (the same for both) has been removed. Use neofs_node_state_epoch for SN and neofs_ir_state_epoch for IR instead.

Deprecated --no-precheck flag in neofs-cli container set-eacl use --force flag instead for skipping validation checks.

Subnets support has been removed:

• IR's workers.subnet and contracts.subnet configs are not used anymore.
• SN's node.subnet config section is not used anymore.
• neoofs-amd morph does not have subnet subcommand anymore.
• neofs-cli container create does not have --subnet flag anymore.

Docker images now contain a single executable file and SSL certificates only.

neofs-cli control healthcheck exit code is 0 only for "READY" state.

To migrate data from Blobovnicza trees to Peapods:

$ blobovnicza-to-peapod -config </path/to/storage/node/config>

For any shard, the data from the configured Blobovnicza tree is copied into a created Peapod file named peapod.db in the directory where the tree is located. For example, /neofs/data/blobovcniza/* -> /neofs/data/peapod.db. Notice that existing Blobovnicza trees are untouched. Configuration is also updated, for example, /etc/neofs/config.yaml -> /etc/neofs/config_peapod.yaml. WARN: carefully review the updated config before using it in the application!

To store small objects in more effective and simple sub-storage Peapod, replace blobovnicza sub-storage with the peapod one in blobstor config section. If storage node already stores some data, don't forget to make data migration described above.

v0.37.0 - Sogado

Added

• neofs_[node|ir]_version with version label (#2326)
• TLS RPC support for IR consensus mode (#2322)
• Multiple neo-go RPC reconnection attempts with a delay (#2307)

Changed

• NeoFS CLI generates random private key automatically if wallet is omitted (#2123)
• morph.validators config can be omitted for IR in local consensus mode but must be non-empty otherwise (#2311)

Fixed

• Inability to restore RPC connection after the second disconnect (#2325)
• Tree service panics when cleaning up failed connections (#2335)
• Dropping small objects on any error on write-cache side (#2336)
• Iterating over just removed files by FSTree (#98)
• IR metrics moved to neofs_ir namespace and epoch metrics to object subsystem (#2344)
• IR metrics config reading (#2344)

Removed

• Non-notary mode support for sidechain (#2321)
• Priority switching b/w RPC endpoints in the morph client (#2306)
• Support for binary keys in the CLI (#2357)

Updated

• Update minimal supported Go version up to v1.18 (#2340)
• tzhash library to 1.7.0 (#2348)
• github.com/hashicorp/golang-lru to v2.0.2
• neofs-sdk-go to v1.0.0-rc.8
• BoltDB (go.etcd.io/bbolt) to 1.3.7

Updating from v0.36.1

• neofs_node_object_epoch metric for IR and SN (the same for both) has been deprecated and will be removed with thenext minor release. Use neofs_node_state_epoch for SN and neofs_ir_state_epoch for IR instead.
• Storage and Inner-ring nodes exposes their version via the neofs_[node|ir]_version metric now.
• In the local consensus mode (IR) it is allowed to provide additional TLS setup addresses now, seemorph.consensus.rpc.tls section.
• morph.switch_interval IR and SN config value is not used anymore.
• morph.rpc_endpoint SN config value and morph.endpoint.client IR config value has been deprecated and will be removed with the next minor release. Use morph.endpoints for both instead (NOTE: it does not have priorities now).
• If you're using binary keys with neofs-cli (-w), convert them to properNEP-6 wallets like this: $ xxd -p < path_to_binary.wallet # outputs hex-encoded key $ neofs-cli util keyer <hex_key> # outputs WIF $ neo-go wallet import -w <wallet_file> --wif <wif_key>or just generate/use new keys.
• In local consensus mode morph.validators in IR's config can be omitted now, morph.consensus.committee will be used instead.For detached consensus, it is a required config parameter now.

v0.36.1

Fixed

• Storage node hanging after RPC disconnect (#2304)
• Using deprecated NeoGo APIs (#2219, #2310)
• Storage node panicking on exit (#2308)
• Inner ring node panic on exit if internal CN is not enabled (#2308)
• Typo in help message (#2240)

Removed

• FallbackTime Morph client configuration option (#2310)

v0.36.0 - Gado

A number of fixes and an internal CN for IR nodes. An update is recommended,data storage scheme is compatible with 0.35.0.

Added

• Doc for extended headers (#2128)
• Separate batching for replicated operations over the same container in pilorama (#1621)
• object.delete.tombstone_lifetime config parameter to set tombstone lifetime in the DELETE service (#2246)
• neofs-adm morph dump-hashes command now also prints NNS domain expiration time (#2295)
• Launch mode of the Inner Ring with local consensus (#2194)

Changed

• common.PrintVerbose prints via cobra.Command.Printf (#1962)
• Storage node's replicator.put_timeout config default to 1m (#2227)
• Full list of container is no longer cached (#2176)
• Pilorama now can merge multiple batches into one (#2231)
• Storage engine now can start even when some shard components are unavailable (#2238)
• neofs-cli buffer for object put increased from 4 KiB to 3 MiB (#2243)
• neofs-adm now reuses network's setting during contract update if they're not overriden by configuration (#2191)

Fixed

• Pretty printer of basic ACL in the NeoFS CLI (#2259)
• Failing SN and IR transactions because of incorrect scopes (#2230, #2263)
• Global scope used for some transactions (#2230, #2263)
• Potential data loss from nodes outside the container or netmap (#2267)
• Invalid Inner Ring listing method through Netmap contract with notary Sidechain (#2283)
• Divide-by-zero panic in Inner Ring's basic income distribution procedure (#2262)
• Big object removal with non-local parts (#1978)
• Disable pilorama when moving to degraded mode (#2197)
• Fetching blobovnicza objects that not found in write-cache (#2206)
• Do not search for the small objects in FSTree (#2206)
• Correct status error for expired session token (#2207)
• Restore subscriptions correctly on morph client switch (#2212)
• Expired objects could be returned if not marked with GC yet (#2213)
• neofs-adm morph dump-hashes now properly iterates over custom domain (#2224)
• Possible deadlock in write-cache (#2239)
• Fix *_req_count and *_req_count_success metric values (#2241)
• Storage ID update by write-cache (#2244)
• neo-go client deadlock on subscription (#2244, #2272)
• Possible panic during write-cache initialization (#2234)
• incorrect NNS resolve handling in neofs-adm (#2296)
• Failed TestNet replication (#2288)

Updated

• neo-go to v0.101.1 (some code also reworked to not use deprecated APIs)
• github.com/klauspost/compress to v1.15.13
• github.com/multiformats/go-multiaddr to v0.8.0
• golang.org/x/term to v0.3.0
• google.golang.org/grpc to v1.51.0
• github.com/nats-io/nats.go to v1.22.1

Updating from v0.35.0

New experimental config field object.delete.tombstone_lifetime allows to settombstone lifetime more appropriate for a specific deployment.

v0.35.0 - Sindo (신도, 信島)

Added

• morph list-containers in neofs-adm (#1689)
• --binary flag in neofs-cli object put/get/delete commands (#1338)
• session flag support to neofs-cli object hash (#2029)
• Shard can now change mode when encountering background disk errors (#2035)
• Background workers and object service now use separate client caches (#2048)
• replicator.pool_size config field to tune replicator pool size (#2049)
• Fix NNS hash parsing in morph client (#2063)
• neofs-cli neofs-cli acl basic/extended print commands (#2012)
• neofs_node_object_*_req_count_success prometheus metrics for tracking successfully executed requests (#1984)
• Metric 'readonly' to get shards mode (#2022)
• Tree service replication timeout (#2159)
• apiclient.reconnect_timeout setting allowing to ignore failed clients for some time (#2164)

Changed

• object lock command reads CID and OID the same way other commands do (#1971)
• LOCK object are stored on every container node (#1502)
• neofs-cli container get-eacl print ACL table in json format only with arg `--json' (#2012)
• Side chain notary deposits use max uint32 as till parameter (#1486)
• Allow object removal without linking object (#2100)
• neofs-cli container delete command pre-checks container ownership (#2106)
• Policer cache size is now 1024 (#2158)
• Tree service now synchronizes with container nodes in a random order (#2127)
• Pilorama no longer tries to apply already applied operations (#2161)
• Use sync.Pool in Object.PUT service (#2139)
• Shard uses metabase for HEAD requests by default, not write-cache (#2167)
• Clarify help for --expire-at parameter for commands object lock/put and bearer create (#2097)
• Node spawns GETRANGE requests signed with the node's key if session key was not found for RANGEHASH (#2144)

Fixed

• Open FSTree in sync mode by default (#1992)
• neofs-cli container nodes's output (#1991)
• Increase error counter for write-cache flush errors (#1818)
• Correctly select the shard for applying tree service operations (#1996)
• Do not panic and return correct errors for bad inputs in GET_RANGE (#2007, #2024)
• Physical child object removal by GC (#1699)
• Broadcasting helper objects (#1972)
• neofs-cli lock object's lifetime flag handling (#1972)
• Do not move write-cache in read-only mode for flushing (#1906)
• Child object collection on CLI side with a bearer token (#2000)
• Fix concurrent map writes in Object.Put service (#2037)
• Malformed request errors' reasons in the responses (#2028)
• Session token's IAT and NBF checks in ACL service (#2028)
• Losing meta information on request forwarding (#2040)
• Assembly process triggered by a request with a bearer token (#2040)
• Losing locking context after metabase resync (#1502)
• Removing all trees by container ID if tree ID is empty in pilorama.Forest.TreeDrop (#1940)
• Concurrent mode changes in the metabase and blobstor (#2057)
• Panic in IR when performing HEAD requests (#2069)
• Write-cache flush duplication (#2074)
• Ignore error if a transaction already exists in a morph client (#2075)
• ObjectID signature output in the CLI (#2104)
• Pack arguments of setPrice invocation during contract update (#2078)
• neofs-cli object hash panic (#2079)
• Closing neo-go WS clients on shutdown and switch processes (#2080)
• Making notary deposits with a zero GAS balance (#2080)
• Notary requests on shutdown (#2075)
• neofs-cli container create check the sufficiency of the number of nodes in the selector for replicas (#2038)
• Data duplication during request forwarding (#2047)
• Tree service panic on TreeMove operation (#2140)
• Panic in GETRANGE with zero length (#2095)
• Spawning useless GETRANGE with zero length for a big object (#2101)
• Incomplete object put errors do contain the deepest error's message (#2092)
• Prioritize internal addresses for clients (#2156)
• Force object removal via control service (#2145)
• Synchronizing a tree now longer reports an error for a single-node container (#2154)
• Prevent leaking goroutines in the tree service (#2162)
• Do not search for LOCK objects when delete container when session provided (#2152)
• Race conditions on shard's mode switch (#1956)
• Returning expired/removed objects from write-cache (#2016)

Removed

• -g option from neofs-cli control ... and neofs-cli container create commands (#2089)
• --header from neofs-cli object get (#2090)

Updated

• neo-go to v0.100.0
• spf13/cobra to v1.6.1
• spf13/viper to v1.8.0
• google.golang.org/grpc to v1.50.1

Updating from v0.34.0

Pass CID and OID parameters via the --cid and --oid flags, not as the command arguments.
Replicator pool size can now be fine-tuned with replicator.pool_size config field.The default value is taken from object.put.pool_size_remote as in earlier versions.
Added neofs_node_object_*_req_count_success metrics for tracking successfully executed requests.
neofs-cli container delete command now requires given account or session issuerto match the container owner. Use --force (-f) flag to bypass this requirement.
Tree service network replication can now be fine-tuned with tree.replication_timeout config field.

v0.34.0 - Marado (마라도, 馬羅島)

Added

• --timeout flag in neofs-cli control commands (#1917)
• Document shard modes of operation (#1909)
• tree list CLI command (#1332)
• TreeService.GetTrees RPC (#1902)
• All trees synchronization on bootstrap (#1902)
• --force flag to neofs-cli control set-status command (#1916)
• Logging SessionService.Create RPC on the server for debug (#1930)
• Debian packages can now be built with make debpackage (#409)

Changed

• Path to a metabase can now be reloaded with a SIGHUP (#1869)

Fixed

• writecache.max_object_size is now correctly handled (#1925)
• Correctly handle setting ONLINE netmap status after maintenance (#1922)
• Correctly reset shard errors in ControlService.SetShardMode RPC (#1931)
• Setting node's network state to MAINTENANCE while network settings forbid it (#1916)
• Do not panic during API client creation (#1936)
• Correctly sign new epoch transaction in neofs-adm for a committee of more than 4 nodes (#1949)
• Inability to provide session to NeoFS CLI in a NeoFS-binary format (#1933)
• neofs-adm now works correctly with a committee of more than 4 nodes (#1949, #1959)
• Closing a shard now waits until GC background workers stop (#1964)
• Make it possible to use shard.ContainerSize in read-only mode (#1975)
• Storage node now starts if at least one gRPC endpoint is available (#1893)
• Panic in API multy client (#1961)
• Blobstor object removal log messages (#1953)
• Missing object relatives in object removal session opened by NeoFS CLI (#1978)
• Bringing a node back online during maintenance (#1900)

Updated

• neo-go to v0.99.4
• protoc to v3.21.7
• neofs-sdk to v1.0.0-rc.7

Updating from v0.33.0

Now storage node serves Control API SetNemapStatus request with MAINTENANCEstatus only if the mode is allowed in the network settings. To force starting the localmaintenance on the node, provide --force flag to the neofs-cli control set-statuscommand.