Proto commits in cs3org/reva

These 91 commits are when the Protocol Buffers files have changed:

Commit:6d592e2
Author:Gianmaria Del Monte
Committer:Gianmaria Del Monte

Merge branch 'cbox-admin-invocations' into cbox-admin-services

Commit:2d56234
Author:Gianmaria Del Monte

Merge branch 'cbox-admin-foundations' into cbox-cli-core

Commit:75d73c6
Author:Gianmaria Del Monte

admin: trim the comments and simplify Assisted-by: Claude:claude-sonnet-5-5

Commit:8e70e11
Author:Gianmaria Del Monte
Committer:Gianmaria Del Monte

admin: trim the comments and simplify Assisted-by: Claude:claude-sonnet-5-5

Commit:08cfe73
Author:Gianmaria Del Monte

Merge branch 'admin-cs3-spaces' into cbox-admin-services

Commit:afaf6c9
Author:Gianmaria Del Monte

admin: SpaceQuota carries what the space uses CS3's Quota has only what remains, which says nothing without a limit or over it, and operators need the usage in both cases. The SpaceQuota wrapper now has used_bytes and used_files from the storage provider; remaining_* in Quota stay computed. The CLI shows USED and FILES from the new fields. Assisted-by: Claude:claude-opus-5-5

Commit:acfa319
Author:Gianmaria Del Monte

admin: SharesAdmin on the CS3 references, filters and updates The requests of SharesAdmin reuse the CS3 messages instead of their own: - a share is a collaboration.ShareReference (id or key), a link a link.PublicShareReference; PublicLinkRef is gone; - a user is a UserId (initiator of the transfers, owner and as of MoveShares and as of ListShares); - ListShares takes repeated collaboration.Filter (OWNER, CREATOR, GRANTEE, RESOURCE_ID, SPACE_ID) and ListPublicLinks repeated ListPublicSharesRequest.Filter (OWNER, CREATOR, RESOURCE_ID), with the admin-only fields; the id of ListShares and the ref of ListPublicLinks are gone, as that is GetShare and GetPublicLink; - UpdateShare takes repeated UpdateShareRequest.UpdateField and UpdatePublicLink repeated UpdatePublicShareRequest.Update, plus fallback. An expiration without a value removes it, which replaces remove_expiration. The facade maps them onto the invocation arguments: permissions become the name of their role, a key becomes an id with `get` before a mutation, and what the invocations cannot do (another filter type, two filters of a type, the display name, the description, the upload notifications) is InvalidArgument. The CLI adapter passes the references and the updates of the user plane through, so ShareUpdate, expirationOf and PublicLinkRef are gone. The operator options of `share list` and `link list` accept --by-resource-id, now a RESOURCE_ID filter. Assisted-by: Claude:claude-opus-5-5

Commit:c3ab8b5
Author:Gianmaria Del Monte
Committer:Gianmaria Del Monte

Merge branch 'admin-cs3-storage' into admin-cs3-spaces The quota target resolves the root of a space with the CS3 Reference that ResolvePath now takes. Assisted-by: Claude:claude-opus-5-5

Commit:34dae24
Author:Gianmaria Del Monte

admin: SpacesAdmin takes and returns the CS3 messages The requests and the responses of SpacesAdmin use the CS3 message of every filter, update, user, space and quota, and add only what CS3 has no field for: - ListSpacesRequest has ListStorageSpacesRequest.Filter (SPACE_TYPE, OWNER, USER, ID, PATH), plus status, profile, include_archived, and created_after/created_before as cs3.types.v1beta1.Timestamp. The list invocation takes one type, owner and member, so the facade keeps the spaces that match an id or a root path. - UpdateSpaceRequest has UpdateStorageSpaceRequest.UpdateField: the description and the readme and thumbnail metadata. A name, an alias, a quota or a field given twice is InvalidArgument. - SetSpaceOwnerRequest.owner and CheckSpacesRequest.owner are a UserId. - RegisterSpaceRequest has a StorageSpace in the form that AdminSpace returns, the roles reader, writer and admin with one group each, plus status. A value that the catalogue would not store is refused. - SpaceQuota has a Quota and a UserId owner. The storage provider reports what is used, so the facade computes what remains, and a limit of 0 is no limit. SetSpaceQuotaRequest has a Quota and an update_mask. The invocations of the spacesregistry and of the storageprovider do not change. The CLI builds the CS3 forms, and shows used as max - remaining. Assisted-by: Claude:claude-opus-5-5

Commit:fe40655
Author:Gianmaria Del Monte

admin: ResolvePath takes a CS3 Reference, CodeFromGRPC in rgrpc/status ResolvePathRequest takes a cs3.storage.provider.v1beta1.Reference, as GetStorageProvidersRequest does. Only an absolute path is supported: the resolve invocation of the storage registry takes only a path, so a reference with a resource id is InvalidArgument. The gRPC to CS3 code table moves to pkg/rgrpc/status as CodeFromGRPC, with its test. It maps OK to CODE_OK; the storage facade turns a failure without a code into Unknown before it maps it. quota.go still builds ResolvePathRequest{Path}: it is adapted at the merge with the SpacesAdmin branch. Assisted-by: Claude:claude-opus-5-5

Commit:f885a3b
Author:Gianmaria Del Monte

admin: StorageAdmin reuses the CS3 ProviderInfo and rpc Status A route and a resolved mount are a cs3.storage.registry.v1beta1.ProviderInfo (provider_path, provider_id, address). ResolvePath wraps it in ResolvedProvider to add the live provider_nodes, which CS3 has no field for. AddRoute takes a ProviderInfo and refuses an address, which the routing table would drop. RemoveRoute keeps the path, the key of the table. A stale instance carries a cs3.rpc.v1beta1.Status whose code is the CS3 form of the gRPC code of the failed reload. The storageregistry invocations do not change: the facade decodes their JSON and maps it. Route and Mount are removed. Assisted-by: Claude:claude-opus-5-5

Commit:a7aa9fa
Author:Gianmaria Del Monte

admin: answer Unimplemented to the profile filter of ListSpaces The catalogue does not store profiles yet. An empty list would say that no space has the profile, which is not known. Assisted-by: Claude:claude-opus-5-5

Commit:e2e20aa
Author:Gianmaria Del Monte

Merge branch 'cbox-admin-invocations' into cbox-admin-services

Commit:faa5a81
Author:Gianmaria Del Monte
Committer:Gianmaria Del Monte

admin: the SharesAdmin service SharesAdmin is the typed operator surface for the shares and public links of any owner, on top of the usershareprovider and publicshareprovider invocations: the fleet-wide lists, get, update and remove for the owner fallback of the CLI, sync of an expired grant, transfer, mark orphan, and MoveShares, which moves the shares and the links under a path after a data move. AdminShare and AdminPublicShare wrap the CS3 messages and add only the orphan state, the current path, password_protected and extra. A list resolves a space given by name through the spacesregistry. MoveShares finds its candidates with the list invocations, because the share tables keep the resource id and not the path. The admin service audits each mutation, with fallback=true when the CLI says so, and never logs a password or a link token. Assisted-by: Claude:claude-opus-5-5

Commit:79f5454
Author:Gianmaria Del Monte
Committer:Gianmaria Del Monte

admin: space quotas on SpacesAdmin GetSpaceQuota, SetSpaceQuota and ListSpaceQuotas run the quota invocations of the storage provider. The facade finds the space in the catalogue, or by its root path or id when it is not there, then the mount that serves the root through the storage registry, then an instance of that mount by its mount_id metadata. The storage chooses the quota form from the space. ListSpaceQuotas asks one instance of each mount, because the instances of a mount share the storage. TestListServicesByInvocation now adds its nodes to the process-wide registry, so that it does not depend on running first. Assisted-by: Claude:claude-opus-5-5

Commit:f1cef1c
Author:Gianmaria Del Monte
Committer:Gianmaria Del Monte

admin: StorageAdmin routes over the storage registry invocations ListRoutes, ResolvePath, AddRoute and RemoveRoute run the invocations of the storage registry. AddRoute and RemoveRoute write once, then reload the table on every instance, and report the instances that did not reload. ResolvePath adds the storage provider instances of each mount, found by their mount_id metadata. The errtypes errors of the registry get their gRPC code. Assisted-by: Claude:claude-opus-5-5

Commit:8fbaff5
Author:Gianmaria Del Monte
Committer:Gianmaria Del Monte

admin: the SpacesAdmin service over the project catalogue SpacesAdmin gives typed RPCs for the space catalogue of the spacesregistry: ListSpaces and CheckSpaces (server streams), GetSpace, UpdateSpace (the server side of the owner fallback), SetSpaceStatus, SetSpaceOwner, RegisterSpace and UnregisterSpace. The profile RPCs are reserved in comments only. Each RPC calls the spacesregistry invocation of the same operation and decodes its result into AdminSpace{space, status, profile, extra}. A SpaceRef by path is resolved in the facade, because the catalogue selects by id or by name only. After a change, flush_cache runs on every instance, because each one caches the project list. The changes are audited as spaces.<invocation>. Assisted-by: Claude:claude-opus-5-5

Commit:5d3894c
Author:Gianmaria Del Monte

invoke: describe the argument types in the catalogue ArgSpec gets type, repeated, default and enum, in pkg/invoke and in the admin and control channel protos. Typed fills them from the argument struct: the field type gives type and repeated, a `default` tag gives the default, and an `enum` tag or a field type with Values() []string gives the enum. bind applies the default of an absent argument and refuses a value outside the enum, so the catalogue tells the truth. A bad default fails when the service starts. The level of the logs invocation is the first enum. Stubs regenerated with `make protobuf CS3APIS_DIR=<cs3apis checkout>`. Assisted-by: Claude:claude-opus-5-5

Commit:4e79286
Author:Gianmaria Del Monte

Merge branch 'impersonation-actor' into cbox-admin-foundations # Conflicts: # internal/grpc/services/admin/impersonate.go # pkg/admin/adminpb/admin.pb.go # pkg/auth/scope/machine.go

Commit:1f407c2
Author:Gianmaria Del Monte

admin: make impersonation safe and visible in the audit Impersonate now: - records the caller in the actor scope of the token (the user id, or socket:<uid> on the local-root socket); - refuses a token that carries an actor, and so does RequestAdmin, so an admin who impersonates another admin cannot step up as that admin; - refuses members of admin_group and service accounts unless the request sets allow_privileged, which is audited on its own line; - gives a lightweight target the scopes of its own login, and refuses account types that cannot log in; - answers NotFound for a target that does not exist; - returns expires_at, and takes an optional ttl_seconds that the server caps at the impersonation lifetime. CheckAdmin reports an impersonated session as not admin. The HTTP endpoint and `reva admin impersonate` get the new options. The CLI asks for confirmation for --allow-privileged. Assisted-by: Claude:claude-opus-5-5

Commit:ee1c092
Author:Gianmaria Del Monte

admin: add the foundation for typed Admin API services Add SpacesAdmin, SharesAdmin and StorageAdmin in their own proto files of reva.admin.v1beta1, still without methods, and the shared messages in types.proto (SpaceRef). Both the gRPC listener and the local socket register every service. The protos import CS3 messages, so make protobuf takes a cs3apis checkout in CS3APIS_DIR. An invocation error now carries a gRPC code through the control channel: invoke.Error, or a status error, sets the new code field of the control and admin results. The admin service gets the helpers that a typed RPC uses on top of the invocations: invokeOne, invokeAll, relayStream and decode. Assisted-by: Claude:claude-opus-5-5

Commit:900b6e0
Author:Gianmaria Del Monte
Committer:GitHub

stats: per-owner share and project statistics (#5943) * stats: per-owner share and project statistics The share managers count the active shares of each creator, and the spaces registry reports the projects, their quota and their used space per owner, so the collector can fold them by the owner attributes of the site. The collector can keep only some attributes as labels and read the owner map from one key of the file. A stats invocation gets 5 minutes instead of the 10 seconds of an admin call, because the spaces registry asks the quota of every project. * spacesregistry: impersonate project owners through the Admin API The project quota is asked as the owner with a token from the Admin API Impersonate call, made as the service spacesregistry, so the spaces registry no longer needs the machine secret and each impersonation is audited. The owner context keeps the deadline of the caller. * stats: declare the stats invocation shared Every Reporter reads state that all instances share, so the invocation is shared and the collector routes it with client.Route and client.Pick: one reachable instance per service instead of the first one. The payload no longer carries a scope, the invocation declares it. * stats: stream the stats through the Admin API The collector no longer reads the registry, mints its own token or routes the calls: it asks ListServices, which gets an invocation filter, for the services with a stats invocation, and runs it with InvokeStream as the service stats-collector. The Admin API already routes a shared invocation to one instance. The invocation is a stream of payloads of at most 10000 samples, because the per-owner samples can pass the gRPC message limit. A stream has no per-instance timeout, so the timeout parameter of InvokeOne is reverted.

Commit:c722517
Author:Gianmaria Del Monte
Committer:GitHub

admin: stream a stream-only invocation without -stream (#5944)

Commit:001860f
Author:Gianmaria Del Monte

admin: stream a stream-only invocation without -stream An invocation spec now says when the invocation has no single-result form (stream_only), and reva admin invoke streams it without -stream. Built-in logs keeps both forms, so a plain invoke still prints a snapshot. A server without the field reports false, which keeps the old behaviour.

Commit:28f6e07
Author:Gianmaria Del Monte

stats: stream the stats through the Admin API The collector no longer reads the registry, mints its own token or routes the calls: it asks ListServices, which gets an invocation filter, for the services with a stats invocation, and runs it with InvokeStream as the service stats-collector. The Admin API already routes a shared invocation to one instance. The invocation is a stream of payloads of at most 10000 samples, because the per-owner samples can pass the gRPC message limit. A stream has no per-instance timeout, so the timeout parameter of InvokeOne is reverted.

Commit:a88c96d
Author:Gianmaria Del Monte
Committer:Jesse Geens

invoke: run shared invocations on one instance An invocation that acts on state all instances share, like the jobs store, ran on every instance: `admin invoke jobs enqueue` enqueued one run per runner. These invocations are now declared Shared() and advertised in the registry, and admin sends them to one live instance per service. It prefers instances in rotation and moves to another one only before the call is sent, so a mutation never runs twice.

Commit:7fc6502
Author:Gianmaria Del Monte
Committer:Jesse Geens

invoke: run shared invocations on one instance An invocation that acts on state all instances share, like the jobs store, ran on every instance: `admin invoke jobs enqueue` enqueued one run per runner. These invocations are now declared Shared() and advertised in the registry, and admin sends them to one live instance per service. It prefers instances in rotation and moves to another one only before the call is sent, so a mutation never runs twice.

Commit:090d711
Author:Gianmaria Del Monte
Committer:Gianmaria Del Monte

invoke: run shared invocations on one instance An invocation that acts on state all instances share, like the jobs store, ran on every instance: `admin invoke jobs enqueue` enqueued one run per runner. These invocations are now declared Shared() and advertised in the registry, and admin sends them to one live instance per service. It prefers instances in rotation and moves to another one only before the call is sent, so a mutation never runs twice.

Commit:bcfa51a
Author:Gianmaria Del Monte
Committer:Gianmaria Del Monte

rjobs: add RetryAfter, Permanent and the aborted state Jobs can choose the retry delay or give up for good. The attempt number now comes from the delivery count.

Commit:4fb8b98
Author:Gianmaria Del Monte
Committer:GitHub

rjobs: add RetryAfter, Permanent and the aborted state Jobs can choose the retry delay or give up for good. The attempt number now comes from the delivery count.

Commit:384d697
Author:Gianmaria Del Monte
Committer:Gianmaria Del Monte

rjobs: let jobs report their progress Jobs implementing ProgressJob get a Reporter. The latest snapshot is persisted periodically and once more before the terminal status, and is shown by the admin API and CLI.

Commit:350723b
Author:Gianmaria Del Monte
Committer:Gianmaria Del Monte

invoke: run shared invocations on one instance An invocation that acts on state all instances share, like the jobs store, ran on every instance: `admin invoke jobs enqueue` enqueued one run per runner. These invocations are now declared Shared() and advertised in the registry, and admin sends them to one live instance per service. It prefers instances in rotation and moves to another one only before the call is sent, so a mutation never runs twice.

Commit:7cd0847
Author:Gianmaria Del Monte
Committer:GitHub

rjobs: add RetryAfter, Permanent and the aborted state Jobs can choose the retry delay or give up for good. The attempt number now comes from the delivery count.

Commit:6a741b4
Author:Gianmaria Del Monte
Committer:GitHub

rjobs: let jobs report their progress Jobs implementing ProgressJob get a Reporter. The latest snapshot is persisted periodically and once more before the terminal status, and is shown by the admin API and CLI.

Commit:8fa2475
Author:Gianmaria Del Monte
Committer:Gianmaria Del Monte

invoke: run shared invocations on one instance An invocation that acts on state all instances share, like the jobs store, ran on every instance: `admin invoke jobs enqueue` enqueued one run per runner. These invocations are now declared Shared() and advertised in the registry, and admin sends them to one live instance per service. It prefers instances in rotation and moves to another one only before the call is sent, so a mutation never runs twice.

Commit:21f011d
Author:Gianmaria Del Monte
Committer:Jesse Geens

rjobs: add RetryAfter, Permanent and the aborted state Jobs can choose the retry delay or give up for good. The attempt number now comes from the delivery count.

Commit:00e13ec
Author:Gianmaria Del Monte
Committer:Jesse Geens

rjobs: let jobs report their progress Jobs implementing ProgressJob get a Reporter. The latest snapshot is persisted periodically and once more before the terminal status, and is shown by the admin API and CLI.

Commit:021fa87
Author:Gianmaria Del Monte
Committer:Gianmaria Del Monte

rjobs: enqueue jobs registered by other runners Runners tag job consumers with metadata, so Enqueue accepts a job that only another process runs. Add 'admin jobs forget' to drop a removed job.

Commit:e30125d
Author:Gianmaria Del Monte
Committer:Gianmaria Del Monte

rjobs: enqueue jobs registered by other runners Runners tag job consumers with metadata, so Enqueue accepts a job that only another process runs. Add 'admin jobs forget' to drop a removed job.

Commit:de50a1a
Author:Gianmaria Del Monte

invoke: run shared invocations on one instance An invocation that acts on state all instances share, like the jobs store, ran on every instance: `admin invoke jobs enqueue` enqueued one run per runner. These invocations are now declared Shared() and advertised in the registry, and admin sends them to one live instance per service. It prefers instances in rotation and moves to another one only before the call is sent, so a mutation never runs twice.

Commit:397c957
Author:Gianmaria Del Monte
Committer:Jesse Geens

Expose admin status and impersonation over HTTP Clients that only speak the public HTTPS surface cannot reach the Admin API, which is gRPC on its own listener. A new admin HTTP service gives them the part they need: GET /admin/status says whether the caller is an admin, and POST /admin/impersonate returns a token acting as another user. Impersonate steps the caller up and impersonates in one request, so the admin token never leaves the server; both steps are audited as before, with a reason if the client gives one. Status is backed by a new CheckAdmin RPC that mints nothing, so asking leaves no audit trail. Impersonation tokens no longer share the admin token's lifetime. An impersonated session is used like any other, long transfers included, and fifteen minutes cut those off. They now last as long as a token from signing in, or impersonation_ttl.

Commit:a1f7696
Author:Gianmaria Del Monte
Committer:GitHub

rjobs: add RetryAfter, Permanent and the aborted state Jobs can choose the retry delay or give up for good. The attempt number now comes from the delivery count.

Commit:2a28aa5
Author:Gianmaria Del Monte
Committer:GitHub

rjobs: let jobs report their progress Jobs implementing ProgressJob get a Reporter. The latest snapshot is persisted periodically and once more before the terminal status, and is shown by the admin API and CLI.

Commit:e4e293f
Author:Gianmaria Del Monte
Committer:GitHub

rjobs: enqueue jobs registered by other runners Runners tag job consumers with metadata, so Enqueue accepts a job that only another process runs. Add 'admin jobs forget' to drop a removed job.

Commit:0998d98
Author:Gianmaria Del Monte
Committer:GitHub

rjobs: add RetryAfter, Permanent and the aborted state Jobs can choose the retry delay or give up for good. The attempt number now comes from the delivery count.

Commit:cc11b07
Author:Gianmaria Del Monte
Committer:GitHub

rjobs: let jobs report their progress Jobs implementing ProgressJob get a Reporter. The latest snapshot is persisted periodically and once more before the terminal status, and is shown by the admin API and CLI.

Commit:c3eac68
Author:Gianmaria Del Monte
Committer:GitHub

rjobs: enqueue jobs registered by other runners Runners tag job consumers with metadata, so Enqueue accepts a job that only another process runs. Add 'admin jobs forget' to drop a removed job.

Commit:73005d6
Author:Gianmaria Del Monte
Committer:Gianmaria Del Monte

rjobs: add RetryAfter, Permanent and the aborted state Jobs can choose the retry delay or give up for good. The attempt number now comes from the delivery count.

Commit:3883e70
Author:Gianmaria Del Monte
Committer:Gianmaria Del Monte

rjobs: let jobs report their progress Jobs implementing ProgressJob get a Reporter. The latest snapshot is persisted periodically and once more before the terminal status, and is shown by the admin API and CLI.

Commit:d366cbf
Author:Gianmaria Del Monte
Committer:Gianmaria Del Monte

rjobs: enqueue jobs registered by other runners Runners tag job consumers with metadata, so Enqueue accepts a job that only another process runs. Add 'admin jobs forget' to drop a removed job.

Commit:594339b
Author:Gianmaria Del Monte
Committer:Gianmaria Del Monte

Expose admin status and impersonation over HTTP Clients that only speak the public HTTPS surface cannot reach the Admin API, which is gRPC on its own listener. A new admin HTTP service gives them the part they need: GET /admin/status says whether the caller is an admin, and POST /admin/impersonate returns a token acting as another user. Impersonate steps the caller up and impersonates in one request, so the admin token never leaves the server; both steps are audited as before, with a reason if the client gives one. Status is backed by a new CheckAdmin RPC that mints nothing, so asking leaves no audit trail. Impersonation tokens no longer share the admin token's lifetime. An impersonated session is used like any other, long transfers included, and fifteen minutes cut those off. They now last as long as a token from signing in, or impersonation_ttl.

Commit:cfe0d1e
Author:Gianmaria Del Monte
Committer:Gianmaria Del Monte

Expose admin status and impersonation over HTTP Clients that only speak the public HTTPS surface cannot reach the Admin API, which is gRPC on its own listener. A new admin HTTP service gives them the part they need: GET /admin/status says whether the caller is an admin, and POST /admin/impersonate returns a token acting as another user. Impersonate steps the caller up and impersonates in one request, so the admin token never leaves the server; both steps are audited as before, with a reason if the client gives one. Status is backed by a new CheckAdmin RPC that mints nothing, so asking leaves no audit trail. Impersonation tokens no longer share the admin token's lifetime. An impersonated session is used like any other, long transfers included, and fifteen minutes cut those off. They now last as long as a token from signing in, or impersonation_ttl.

Commit:f75964b
Author:Gianmaria Del Monte

Expose admin status and impersonation over HTTP Clients that only speak the public HTTPS surface cannot reach the Admin API, which is gRPC on its own listener. A new admin HTTP service gives them the part they need: GET /admin/status says whether the caller is an admin, and POST /admin/impersonate returns a token acting as another user. Impersonate steps the caller up and impersonates in one request, so the admin token never leaves the server; both steps are audited as before, with a reason if the client gives one. Status is backed by a new CheckAdmin RPC that mints nothing, so asking leaves no audit trail. Impersonation tokens no longer share the admin token's lifetime. An impersonated session is used like any other, long transfers included, and fifteen minutes cut those off. They now last as long as a token from signing in, or impersonation_ttl.

Commit:a786472
Author:Jesse Geens
Committer:GitHub

Feature/admin api (#5773) * Add the Admin API and control channel protobuf definitions * Add the invocation framework, admin scope and registry selection support * Host a per-process control channel and register service instances by node id * Add the admin gRPC service, its loader wiring and the reva admin CLI * Document the Admin API and add its changelog * Serve admin local root over a Unix socket authenticated by OS credentials * Register serverless services on the control channel by control-address node id * Buffer recent log lines in memory for the Admin API to read * Add streaming invocations and the control InvokeStream RPC * Expose recent logs as a built-in service invocation * Read and follow service logs with reva admin logs * Tag request logs with their service and user so admin logs finds them * Trace a request or a user across the fleet with reva admin trace * Report build information through a built-in version invocation * Dump goroutine stacks with reva admin stack * Report and set a process's log level at runtime with reva admin logs level * Diff a service's config across instances with reva admin config -diff * Move the reva admin CLI into its own cmd/reva/admin package * Reset a service's log level back to its configured value with reva admin logs level reset * Take instances out of rotation at runtime with reva admin services drain/enable * Skip the user-group lookup for admin tokens so local-root fan-out works under skip_user_groups_in_token * Report in-flight request activity so an operator knows when a drained instance has quiesced * Give admin subcommands a real -h synopsis and services a grouped subcommand guide * Interact with the jobs runner through typed reva admin jobs API methods over the invoke channel * Keep a job run's start time on its terminal status write * Report the reva version in the version invocation * Fix protolint issues in the admin and control protobufs * Updated doc * Added further clarification comments --------- Co-authored-by: Gianmaria Del Monte <g.macmount@gmail.com> Co-authored-by: Giuseppe Lo Presti <giuseppe.lopresti@cern.ch>

Commit:6e6d92a
Author:Gianmaria Del Monte
Committer:Gianmaria Del Monte

Fix protolint issues in the admin and control protobufs

Commit:3e5d3cc
Author:Gianmaria Del Monte
Committer:Gianmaria Del Monte

Interact with the jobs runner through typed reva admin jobs API methods over the invoke channel

Commit:0039085
Author:Gianmaria Del Monte
Committer:Gianmaria Del Monte

Read and follow service logs with reva admin logs

Commit:8088e9c
Author:Gianmaria Del Monte
Committer:Gianmaria Del Monte

Add streaming invocations and the control InvokeStream RPC

Commit:b8f1c69
Author:Gianmaria Del Monte
Committer:Gianmaria Del Monte

Add the Admin API and control channel protobuf definitions

Commit:5a88060
Author:Gianmaria Del Monte
Committer:GitHub

Fix protolint issues in the admin and control protobufs

Commit:02fb9a7
Author:Gianmaria Del Monte
Committer:GitHub

Interact with the jobs runner through typed reva admin jobs API methods over the invoke channel

Commit:f048915
Author:Gianmaria Del Monte
Committer:GitHub

Read and follow service logs with reva admin logs

Commit:5762522
Author:Gianmaria Del Monte
Committer:GitHub

Add streaming invocations and the control InvokeStream RPC

Commit:4ef4aff
Author:Gianmaria Del Monte
Committer:GitHub

Add the Admin API and control channel protobuf definitions

Commit:c769a03
Author:Gianmaria Del Monte
Committer:Jesse Geens

Fix protolint issues in the admin and control protobufs

Commit:ce454cf
Author:Gianmaria Del Monte
Committer:Jesse Geens

Interact with the jobs runner through typed reva admin jobs API methods over the invoke channel

Commit:92e49d2
Author:Gianmaria Del Monte
Committer:Jesse Geens

Read and follow service logs with reva admin logs

Commit:f7b152e
Author:Gianmaria Del Monte
Committer:Jesse Geens

Add streaming invocations and the control InvokeStream RPC

Commit:b318d63
Author:Gianmaria Del Monte
Committer:Jesse Geens

Add the Admin API and control channel protobuf definitions

Commit:8d1b4ab
Author:Giuseppe Lo Presti
Committer:GitHub

Pass lock metadata on uploads + use upstream EOS GRPC bindings (#4514)

The documentation is generated from this commit.

Commit:d3d77c7
Author:Giuseppe Lo Presti

Updated copyright

Commit:084c1b5
Author:Hugo Labrador
Committer:GitHub

New observability + general refactor of http/grpc clients (#4166) * remove unused fs and tracing * add helloworld example and fix auth logic to bail out early * remove hack on appctx and add traceid * configurable prom collectors * grpc reflection working * wire http traffic * intrument outgoing grpc calls * add stuff * merge appctx and ctx * refactor rhttp to httpclient * tidy go.mod * remove drone tests * add changelog * fix test * go 1.21 * pass tests * fix sql tests * go fmt * more linters * make all linters happy * update docs * add back ceph example * remove otelhttp library * remove files and add tests * align with upstream

Commit:605cec4
Author:Miroslav Bauer
Committer:GitHub

Bump the Copyright date to 2023 (#3584) * Bump the Copyright date to 2023

Commit:1e612bc
Author:Gianmaria Del Monte
Committer:jkoberg

EOS grpc fixes (#3420) * return inode in fileinfo * fix log * fix eos grpc parsing * stat as a root * workaround for inode (revert me) * set parent id in eosfs * Revert "stat as a root" This reverts commit 81ce7b3e66f79d2401549d8ec67c1fa15a187f7b. * set storage id to parent id in storage provider * fix file path * clean path returned by eos * set xattrs for lock as root * fix key when setting attribute * stat as root * Revert "stat as root" This reverts commit e865e6e2f8d75d3759f2e3e0c8ca657a2c6ade6b. * trim prefix for user attrs * do not filter sys attrs * fix add acl * fix initiate file upload reference resolution * fix initiate file upload * Revert "fix initiate file upload" This reverts commit 265d09187d3ab4d646ace20df2ceaf925788e39f. * set sys acl as root * workaround for adding acl * fix unset attr * unset lock as root * Avoid crash when getting checksum in a stat request * fix * add changelog * fix linter * fix license Co-authored-by: Fabrizio Furano <furano@cern.ch>

Commit:f4ad966
Author:Gianmaria Del Monte
Committer:jkoberg

Implementation of CS3 Lock API for EOS storage driver (#2444)

Commit:48a2870
Author:Gianmaria Del Monte
Committer:GitHub

EOS grpc fixes (#3420) * return inode in fileinfo * fix log * fix eos grpc parsing * stat as a root * workaround for inode (revert me) * set parent id in eosfs * Revert "stat as a root" This reverts commit 81ce7b3e66f79d2401549d8ec67c1fa15a187f7b. * set storage id to parent id in storage provider * fix file path * clean path returned by eos * set xattrs for lock as root * fix key when setting attribute * stat as root * Revert "stat as root" This reverts commit e865e6e2f8d75d3759f2e3e0c8ca657a2c6ade6b. * trim prefix for user attrs * do not filter sys attrs * fix add acl * fix initiate file upload reference resolution * fix initiate file upload * Revert "fix initiate file upload" This reverts commit 265d09187d3ab4d646ace20df2ceaf925788e39f. * set sys acl as root * workaround for adding acl * fix unset attr * unset lock as root * Avoid crash when getting checksum in a stat request * fix * add changelog * fix linter * fix license Co-authored-by: Fabrizio Furano <furano@cern.ch>

Commit:af8fe04
Author:Vasco Guita
Committer:Gianmaria Del Monte

Migrate the BuildOnly job from Drone to GitHub Actions

Commit:d25a3a7
Author:Gianmaria Del Monte
Committer:GitHub

Implementation of CS3 Lock API for EOS storage driver (#2444)

Commit:2ec4ae6
Author:Fabrizio Furano
Committer:GitHub

Implement rollback to version (#1927)

Commit:c95ddaf
Author:Ishank Arora
Committer:GitHub

Generate updated protobuf bindings for EOS GRPC (#1916)

Commit:692f38f
Author:Fabrizio Furano
Committer:GitHub

EOS GRPC interface (#1471)

Commit:73f1c7b
Author:Ishank Arora
Committer:GitHub

Add FindAcceptedUsers method to OCM Invite API (#1527)

Commit:bd7234e
Author:Samuel Alfageme
Committer:GitHub

Bump the Copyright in the header to 2021 (#1397)

Commit:655c9ba
Author:Fabrizio Furano
Committer:GitHub

Progress with the eosgrpc client (#1154) Co-authored-by: Ishank Arora <ishank011@gmail.com>

Commit:07ab188
Author:Phil Davis
Committer:GitHub

Change percentagused to percentageused (#903)

Commit:e8f1815
Author:Phil Davis
Committer:GitHub

Fix minor typos (#898)

Commit:0e74029
Author:Fabrizio Furano
Committer:GitHub

eos grpc driver (#664)

Commit:0dbce24
Author:Hugo G. Labrador
Committer:GitHub

Update license year to 2020 (#504)

Commit:f3c5618
Author:Hugo G. Labrador
Committer:GitHub

build: remove vendor modules (#465)

Commit:02e4953
Author:Hugo G. Labrador
Committer:GitHub

vendor deps for network isolated builds (#308) * vendor deps for network isolated builds * clean build * improve ci * speed up build

Commit:9db3db0
Author:Hugo G. Labrador
Committer:GitHub

Gateway+metadata (#289) * Add metadata support * Add gateway as required element of the deployment * New code layout enforcing Go best practices for encapsulation * Support for multiple authentication backends * New user provider grpc service