Introducing vault sharing and the restrictions of pulling and cloning with vault permissions #266

CMCDragonkai · 2021-10-25T13:28:44Z

Description

This pull request aims to allow vaults to be shared and unshared with other Node Ids. Specification and discussion can be found inside the Tasklist issues.

Secret Commands

There are some changes to secrets commands here. These should be extracted out of this PR and moved to another PR as they have different specification requirements.

Issues Fixed

Fixes Complete shareVault/unshareVault implementation #259
Fixes No longer preserve vault ID on cloneVault #253
Fixes Redesign usage of CDSS with VaultInternal #305
Fixes Review js-id usage for NodeId, VaultId, NotificationId, PermId, ClaimId, Gestalts and GenericIdTypes.ts #299
Fixes Investigate usage of the remote field for vault metadata optimisation #309
Fixes Vault Mutation and Reading #335
Fixes VaultInternal Version and Logging #336
Fixes VaultId and the removal of GenericIdTypes #337
Fixes Integration of NodeConnectionManager for Vaults domain #338
Fixes Vault Scanning #339
Fixes Vaults general fixes and QOL #340
Fixes Agent service utility to acquire trusted ConnectionInfo from connecting client node #342
Related to Fine Grained Concurrency Control Standardisation #294 - in terms of introducing withF and withG to be used
Related to Create diagrams for new processes implemented in vaults refactoring MR Polykey-Docs#3 - review sharing diagrams when changes are done

Tasks

Reference for residual TODOs is here #266 (comment) .

Final checklist

CMCDragonkai · 2021-10-25T13:30:01Z

This will require squashing before rebase on master.

scottmmorris · 2021-10-26T06:12:15Z

I'm going to remove the src/vaults/old folder in this MR along with the GitRequest stuff

scottmmorris · 2021-10-26T06:37:46Z

I'm also going to handle the scan vaults function here. To see more info on this see: #259 (comment)

scottmmorris · 2021-10-26T08:17:59Z

vault scan has been reintroduced and some testing has been done but there still needs to be fixes for the CLI and then the testing of the CLI commands.

After discussing with Roger we now have a solution to handle the errors coming through isomorphic git. More details can be found in #259. So I've updated the task list with the relevant tasks.

After those are done this PR should be ready for review. Probably wont finish it this week because ILL only be doing a half day Friday but should be ready by end of Monday next week.

scottmmorris · 2021-10-29T01:44:09Z

I'm getting a connection warning when testing the scan vaults function on node manager:

WARN:Connection 127.0.0.1:41241:Connection Error: Error: This socket has been ended by the other party
WARN:Connection 0.0.0.0:56613:Connection Error: Error [ERR_STREAM_WRITE_AFTER_END]: write after end

I haven't been able to trace it yet but it doesn't seem to be affecting the output from the grpc stream as the tests are still passing. I believe the first error message shouldn't have the '127.0.0.1' address, it should be the default '0.0.0.0' so maybe this is contributing to the issue.

scottmmorris · 2021-10-29T04:44:46Z

@CMCDragonkai

I'm struggling a little with the error handling. At the moment this is what I'm doing. On the server side I am generating and throwing the error and returning.

genWritable.throw(new vaultsErrors.ErrorVaultPermissionDenied);
return;

Then on the client side inside the request object I catch the error from the stream (there is a bit of a hack I'm using to get the what the error is because its coming out weird), set the error flag and then throw the error again.

response.stream.on('error', (err) => {
  // set the error flag
  // throw the error
});

Then I catch the error in the git clone command and do some checks to verify and then throw the correct error.

      try {
        await git.clone({
          ...
        });
      } catch (err) {
        if (err instanceof git.Errors.SmartHttpError && error) {
          if (error === 'ErrorVaultPermissionDenied') throw new vaultsErrors.ErrorVaultPermissionDenied;
        }
        throw err;
      }

However, the issue is that its not executing the try catch wrapper for the git clone method until after my test is finished.

  ●  Cannot log after tests are done. Did you forget to wait for something async in your test?
    Attempted to log "ErrorVaultPermissionDenied: 
        at /home/scottmorris/Documents/js-polykey/src/vaults/VaultManager.ts:495:61
        at processTicksAndRejections (internal/process/task_queues.js:95:5)
        at Object._transaction (/home/scottmorris/Documents/js-polykey/src/vaults/VaultManager.ts:198:14)
        at Object.cloneVault (/home/scottmorris/Documents/js-polykey/src/vaults/VaultManager.ts:371:12)
        at Object.<anonymous> (/home/scottmorris/Documents/js-polykey/tests/vaults/VaultManager.test.ts:867:11) {
      description: 'ErrorVaultPermissionDenied: Polykey error',
      exitCode: 1,
      data: {}
    }".

      493 |         // isomorphic git then we need to throw the polykey error
      494 |         if (err instanceof git.Errors.SmartHttpError && error) {
    > 495 |           if (error === 'ErrorVaultPermissionDenied') throw new vaultsErrors.ErrorVaultPermissionDenied;
          |                                                             ^
      496 |         }
      497 |         throw err;
      498 |       }

It appears that the test shuts down after the stream error is caught and rethrown. So instead of failing with the exception in the try catch its failing within the request object and I assume it isn't being caught there?

EDIT: The alternative I have found is that if I don't throw an error to isomorphic git then it will continue until it realises its not being sent any data from the stream and then will throw and error which is correctly caught and handled. I'm not too sure why there is a difference between these two approaches? By my reckoning they should both cause the process to exit gracefully.

CMCDragonkai · 2021-10-29T06:01:56Z

Are you meant to await on the throw? This usually means somewhere an await is missing.

scottmmorris · 2021-10-29T06:22:52Z

Yep, good point. Although I've changed it and i'm still encountering the async error.

I tried stream.emit as well before as well and it didnt work and that isnt async.

scottmmorris · 2021-11-01T04:24:23Z

I'm now running a try and catch wrapper like this:

              const response = client.vaultsGitInfoGet(request, meta);
              // The polykey error is caught and the error flag set
              try {
                for await (const resp of response) {
                  yield resp.getChunk_asU8();
                }
              } catch (err) {
                error = err;
                throw err;
              }

and it is working fine as opposed to using the response.stream.on('error').

scottmmorris · 2021-11-01T04:26:30Z

The only thing is, that isomorphic git is logging out an error here: https://github.com/isomorphic-git/isomorphic-git/blob/89c0da78d5ebf3c9f2754b3c8d557155dd70c8d7/src/models/GitPktLine.js#L89

Which is this error

    error TypeError: Cannot read property 'length' of undefined
        at StreamReader.read (/home/scottmorris/Documents/js-polykey/node_modules/isomorphic-git/index.cjs:2299:39)
        at processTicksAndRejections (internal/process/task_queues.js:95:5)
        at read (/home/scottmorris/Documents/js-polykey/node_modules/isomorphic-git/index.cjs:6374:22)
        at parseRefsAdResponse (/home/scottmorris/Documents/js-polykey/node_modules/isomorphic-git/index.cjs:6425:17)
        at Function.discover (/home/scottmorris/Documents/js-polykey/node_modules/isomorphic-git/index.cjs:6625:28)
        at _fetch (/home/scottmorris/Documents/js-polykey/node_modules/isomorphic-git/index.cjs:7257:22)
        at _clone (/home/scottmorris/Documents/js-polykey/node_modules/isomorphic-git/index.cjs:7634:42)
        at Object.clone (/home/scottmorris/Documents/js-polykey/node_modules/isomorphic-git/index.cjs:7757:12)
        at /home/scottmorris/Documents/js-polykey/src/vaults/VaultManager.ts:475:9
        at Object._transaction (/home/scottmorris/Documents/js-polykey/src/vaults/VaultManager.ts:198:14)

I think this is because when we throw an error the buffers collected from the generator are undefined

CMCDragonkai · 2021-11-01T04:28:19Z

That's the correct way to handle errors when using the promise/generator utilities.

The only place we are using on event handling is for metadata, and soon that should be turned into a promise too.

scottmmorris · 2021-11-01T05:03:29Z

The only thing is, that isomorphic git is logging out an error here: https://github.com/isomorphic-git/isomorphic-git/blob/89c0da78d5ebf3c9f2754b3c8d557155dd70c8d7/src/models/GitPktLine.js#L89

Which is this error

    error TypeError: Cannot read property 'length' of undefined
        at StreamReader.read (/home/scottmorris/Documents/js-polykey/node_modules/isomorphic-git/index.cjs:2299:39)
        at processTicksAndRejections (internal/process/task_queues.js:95:5)
        at read (/home/scottmorris/Documents/js-polykey/node_modules/isomorphic-git/index.cjs:6374:22)
        at parseRefsAdResponse (/home/scottmorris/Documents/js-polykey/node_modules/isomorphic-git/index.cjs:6425:17)
        at Function.discover (/home/scottmorris/Documents/js-polykey/node_modules/isomorphic-git/index.cjs:6625:28)
        at _fetch (/home/scottmorris/Documents/js-polykey/node_modules/isomorphic-git/index.cjs:7257:22)
        at _clone (/home/scottmorris/Documents/js-polykey/node_modules/isomorphic-git/index.cjs:7634:42)
        at Object.clone (/home/scottmorris/Documents/js-polykey/node_modules/isomorphic-git/index.cjs:7757:12)
        at /home/scottmorris/Documents/js-polykey/src/vaults/VaultManager.ts:475:9
        at Object._transaction (/home/scottmorris/Documents/js-polykey/src/vaults/VaultManager.ts:198:14)

I think this is because when we throw an error the buffers collected from the generator are undefined

Maybe this is because they use a stream reader, and this error occurs essentially when initialising the stream reader because there are no buffers sent by the generator. So then it doesn't have a length and gives this error. So isomorphic git isn't actually reacting to the error we are throwing in the request object. I tried sending some buffers before throwing but the error still occurs.

I then moved onto this structure:

              try {
                for await (const resp of response) {
                  yield resp.getChunk_asU8();
                }
              } catch (err) {
                error = err;
                throw err;
              } finally {
                return {
                  url: url,
                  method: method,
                  body: [],
                  headers: headers,
                  statusCode: 2,
                  statusMessage: 'NOT OK',
                };
              }

so that isomorhic git knows ahead of time that there is an error. This fixes the logging of the type error. I'm not sure if there is a convention I should be using for the status code and message here.

E: actually this doesn't fix it, for some reason it doesn't actually receive the return object in the finally statement

CMCDragonkai · 2021-11-01T06:32:46Z

I think you're translating GRPC exceptions/errors straight into HTTP errors there. This is definitely doable. HTTP has both error status codes as as well as status messages. https://en.wikipedia.org/wiki/List_of_HTTP_status_codes#4xx_client_errors

The question is what does isogit do with this error information? Does it even throw it or it's just dropped. I think the snippet you showed before just dropped the error. In which case these codes aren't even really read.

Note that our exceptions do have a exitCode and description, that could be used if you want some default values to be returned. See src/errors.ts ErrorPolykey.

scottmmorris · 2021-11-01T07:14:56Z

They do check in some places, e.g. here https://github.com/isomorphic-git/isomorphic-git/blob/89c0da78d5ebf3c9f2754b3c8d557155dd70c8d7/src/managers/GitRemoteHTTP.js#L131

However, when I log out the response on the isomorphic git side, it always comes out with the status 200.

Anyway, going back to the original issue which was the iso git error being logged out, the reason its logging out is because the thrown error resulted in no buffers being collected as I mentioned in the previous post. And I think the pk error is only thrown when the buffers are read in from the generator. So in the finally statement I just sent an empty buffer which meant that the length error wasn't thrown but the pk error was.

E: The return statement in the finally block overwrites the error thrown to iso-git but instead iso-git throws another error (emptyServerResponse). Because iso-git generalises errors here to SmartHttpErrors it doesn't really matter what the error is does it? Also eslint complains about the return in the finally block but I have disabled for now. Maybe in the future if what I'm doing is ok then we can remove the throw in the catch statement and just return an empty response instead.

scottmmorris · 2021-11-02T06:37:44Z

Update here: The rebase has caused quite a few test failures. The ones that I'm still struggling with are the agent ones. This is probably due to the fact that fresh wasn't really being used to start the polykey agent but required when starting the vault manager due to the vault key now being stored there

CMCDragonkai · 2021-11-02T07:18:54Z

The API change mostly affected src/proto/schemas. All other changes elsewhere should have been just import path changes and renaming of the protobuf types being used. Although this is subject more change from: #249 (comment)

So it shouldn't be causing new test failures. I did run most of the tests prior to merging including the vaults domain. So not sure why fresh should be involved here.

scottmmorris · 2021-11-04T23:26:33Z

I think its because of the change to the vault key being stored inside vault manager now (So not the rebase changes). I was doing it so that if fresh was specified in the vault manager it would create the key but otherwise it should read the key. But in the keys domain when it was setup it would check if the key exists and then if it didnt it would create it, regardless of whether fresh was specified or not (hence why there were no issues before).

Now that Ive correctly updated it everywhere it should be working. Also there wasn't an option to start the polykey agent from scratch (fresh)? I'm not sure if this is intentional and if then I should just do the vault key the way it was done before. i.e. not rely on the fresh option and instead just check if the key exists and if not create it. I feel like this would run into a lot of problems though.

CMCDragonkai · 2021-11-05T01:17:34Z

@scottmmorris after you finish testing this. Can you go to MatrixAI/Polykey-Docs#3 and update the vault architecture diagrams and the sharing/unsharing sequence diagrams in relation the work done here. Just make a comment there regarding updates to the diagrams, and then coordinate with @joshuakarp in terms of fixing up the wiki.

src/agent/utils.ts

src/bin/agent/start.ts

src/vaults/VaultInternal.ts

- NodeId fixes for vaults tests - Replaced makeVaultIdPretty` with `encodeVaultId`

tegefaulkes · 2022-02-28T01:38:00Z

I have re-based on master. I discovered the interactive re-base feature so I cleaned up the commit history into single commits for each issue as well.

CMCDragonkai · 2022-02-28T08:28:59Z

@tegefaulkes great rebasing however please write descriptive commit messages. You also don't need to reference the issue if it's part of a MR/PR that is committed together.

tegefaulkes · 2022-02-28T23:45:01Z

I'll add better descriptions but I'm going to keep the issue numbers in the commits. This is just so I can track what each change relates to better and when it comes to the last stage rebase-merge step I can squash each issue together nicely. anything labelled fixes for #X are non-specific fixes that will be squashed at some point.

tegefaulkes · 2022-03-01T06:00:00Z

MatrixAI/Polykey-CLI#80 will likely be moved outside of the PR into it's own PR. The changes are not required here and don't fix anything. So to keep things neat it will be seperate.

tegefaulkes · 2022-03-03T01:54:36Z

On review the remaining issues It looks like the rest of the implementation should take 2-3 days. fixing and updating tests should be 2-3 days aswell. Review will possibly add a day or two on top. I'm expecting about to be done halfway through next week, Possibly the end of next week.

tegefaulkes · 2022-03-08T06:55:32Z

All tests are passing now. all that is left is a small review of #336 and an overall review of the code before squashing and merging.

tegefaulkes · 2022-03-08T07:10:19Z

Only one test failing in pipeline. It's working in my tests so i'll take a closer look.

- `VaultId` usage, - updated validation - removal of `GenericIdTypes.ts`

- selective imports - error descriptions updated - Using google `Timestamp` message in `LogEntry` message - fixing comments

…ltsPermissionGet` and `vaultsPermissionUnset`.

- prototyping committing history - prototyping committing, EOD - Linear and branching committing implemented - recovery from dirty state and expanding tests - added global garbage collection for git objects.

CMCDragonkai · 2022-03-10T10:33:32Z

A few things I found that seems to be missing from this PR:

The GRPC tests for vaults have not yet been split like the other GRPC tests. See tests/client/rpcVaults.test.ts. These should go into the tests/service calls. - Split Secrets and Vault Command Tests into Separate Files Polykey-CLI#80
The bin CLI tests for vaults have not been split, see tests/bin/vaults/vaults.test.ts. These should be split into their individual commands like the other tests. - Split Secrets and Vault Command Tests into Separate Files Polykey-CLI#80
The usage of jest.mock is @/keys/utils should be replaced with the usage of the setupGlobalKeypair utility.
Vault ops and secrets-related commands review will be a large epic before we can provide general usage.

@tegefaulkes can you make sure that 1 - 3 have issues tracking them? @emmacasolin can deal with the test splitting soon.

tegefaulkes · 2022-03-11T00:08:35Z

Note that for 3, that some of the tests require different NodeIds to work. This is because we are trying pull, clone or verify permissions in some tests.

As for 4, can you elaborate on the scope and end goal of the issue?

CMCDragonkai · 2022-03-11T01:13:05Z

Note that for 3, that some of the tests require different NodeIds to work. This is because we are trying pull, clone or verify permissions in some tests.

As for 4, can you elaborate on the scope and end goal of the issue?

Ok in that case it's just a matter of create new agents with different keys. It depends on the test abstraction level. Can be rolled into solving 1. and 2.
This requires a overall review of the commands, there are some existing issues in this area but I haven't had a deep look yet. Can you link them into 1 epic? I think there were some things about matching the Unix command API.

CMCDragonkai · 2022-03-11T07:01:55Z

I think MatrixAI/Polykey-CLI#32 can be expanded/elaborated to incorporate point 4. similar to MatrixAI/Polykey-CLI#6.

CMCDragonkai assigned scottmmorris Oct 25, 2021

scottmmorris force-pushed the vaultspermissions branch from f63d1c6 to 354cc77 Compare October 26, 2021 03:03

scottmmorris mentioned this pull request Oct 26, 2021

Complete shareVault/unshareVault implementation #259

Closed

5 tasks

CMCDragonkai requested a review from joshuakarp October 27, 2021 04:04

CMCDragonkai mentioned this pull request Nov 1, 2021

Protobuf domains #280

Merged

11 tasks

scottmmorris force-pushed the vaultspermissions branch from 7f46a63 to 65bdeee Compare November 2, 2021 01:44

CMCDragonkai mentioned this pull request Nov 2, 2021

Implementing BIP39 recovery, Unattended Bootstrap and Agent Start and Status #283

Merged

43 tasks

scottmmorris reviewed Nov 5, 2021

View reviewed changes

src/agent/utils.ts Outdated Show resolved Hide resolved

scottmmorris reviewed Nov 5, 2021

View reviewed changes

src/bin/agent/start.ts Outdated Show resolved Hide resolved

scottmmorris reviewed Nov 5, 2021

View reviewed changes

src/vaults/VaultInternal.ts Outdated Show resolved Hide resolved

scottmmorris force-pushed the vaultspermissions branch 2 times, most recently from 8872ffd to 6d9c3cc Compare November 11, 2021 01:30

Vault sharing, permissions and scanning:

204d3b2

- NodeId fixes for vaults tests - Replaced makeVaultIdPretty` with `encodeVaultId`

tegefaulkes force-pushed the vaultspermissions branch from 17adca0 to 0f03127 Compare February 28, 2022 01:36

tegefaulkes and others added 9 commits March 9, 2022 19:11

Fixes #342 - added connectionInfoGetter utility

202c444

Fixes #305

6c1a4ce

Fixes #337

3a5a34b

- `VaultId` usage, - updated validation - removal of `GenericIdTypes.ts`

Fixes #340

96c752e

- selective imports - error descriptions updated - Using google `Timestamp` message in `LogEntry` message - fixing comments

Fixes #339 - Vaults scanning

409aeea

Fixes #259 replacing un/shareVault with vaultsPermissionSet, `vau…

10cbcce

…ltsPermissionGet` and `vaultsPermissionUnset`.

Fixes #335

205ce25

- prototyping committing history - prototyping committing, EOD - Linear and branching committing implemented - recovery from dirty state and expanding tests - added global garbage collection for git objects.

Fixes #336

4f45490

General fixes

08f6c00

tegefaulkes force-pushed the vaultspermissions branch from 23f2b27 to 08f6c00 Compare March 9, 2022 08:35

tegefaulkes mentioned this pull request Mar 9, 2022

Align VaultInternal with context API functions with NodeConnectionManager #356

Closed

tegefaulkes marked this pull request as ready for review March 9, 2022 09:10

tegefaulkes merged commit 9983ae7 into master Mar 9, 2022

tegefaulkes deleted the vaultspermissions branch March 9, 2022 09:12

tegefaulkes mentioned this pull request Mar 11, 2022

Prevent checking out discarded vault commit IDs with version #252

Closed

CMCDragonkai mentioned this pull request Mar 11, 2022

Seed node not adding details of connecting node to its NodeGraph #344

Closed

6 tasks

CMCDragonkai mentioned this pull request Mar 13, 2022

Incorporate Unix Commands to Secrets Subcommand - pk secrets * MatrixAI/Polykey-CLI#32

Closed

This pull request was closed.

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Introducing vault sharing and the restrictions of pulling and cloning with vault permissions #266

Introducing vault sharing and the restrictions of pulling and cloning with vault permissions #266

CMCDragonkai commented Oct 25, 2021 •

edited by tegefaulkes

Loading

CMCDragonkai commented Oct 25, 2021

scottmmorris commented Oct 26, 2021

scottmmorris commented Oct 26, 2021

scottmmorris commented Oct 26, 2021

scottmmorris commented Oct 29, 2021

scottmmorris commented Oct 29, 2021 •

edited

Loading

CMCDragonkai commented Oct 29, 2021 via email •

edited

Loading

scottmmorris commented Oct 29, 2021

scottmmorris commented Nov 1, 2021

scottmmorris commented Nov 1, 2021

CMCDragonkai commented Nov 1, 2021

scottmmorris commented Nov 1, 2021 •

edited

Loading

CMCDragonkai commented Nov 1, 2021

scottmmorris commented Nov 1, 2021 •

edited

Loading

scottmmorris commented Nov 2, 2021

CMCDragonkai commented Nov 2, 2021

scottmmorris commented Nov 4, 2021 •

edited

Loading

CMCDragonkai commented Nov 5, 2021

tegefaulkes commented Feb 28, 2022

CMCDragonkai commented Feb 28, 2022

tegefaulkes commented Feb 28, 2022 •

edited

Loading

tegefaulkes commented Mar 1, 2022

tegefaulkes commented Mar 3, 2022

tegefaulkes commented Mar 8, 2022

tegefaulkes commented Mar 8, 2022

CMCDragonkai commented Mar 10, 2022 •

edited

Loading

tegefaulkes commented Mar 11, 2022

CMCDragonkai commented Mar 11, 2022

CMCDragonkai commented Mar 11, 2022

Introducing vault sharing and the restrictions of pulling and cloning with vault permissions #266

Introducing vault sharing and the restrictions of pulling and cloning with vault permissions #266

Conversation

CMCDragonkai commented Oct 25, 2021 • edited by tegefaulkes Loading

Description

Secret Commands

Issues Fixed

Tasks

Final checklist

CMCDragonkai commented Oct 25, 2021

scottmmorris commented Oct 26, 2021

scottmmorris commented Oct 26, 2021

scottmmorris commented Oct 26, 2021

scottmmorris commented Oct 29, 2021

scottmmorris commented Oct 29, 2021 • edited Loading

CMCDragonkai commented Oct 29, 2021 via email • edited Loading

scottmmorris commented Oct 29, 2021

scottmmorris commented Nov 1, 2021

scottmmorris commented Nov 1, 2021

CMCDragonkai commented Nov 1, 2021

scottmmorris commented Nov 1, 2021 • edited Loading

CMCDragonkai commented Nov 1, 2021

scottmmorris commented Nov 1, 2021 • edited Loading

scottmmorris commented Nov 2, 2021

CMCDragonkai commented Nov 2, 2021

scottmmorris commented Nov 4, 2021 • edited Loading

CMCDragonkai commented Nov 5, 2021

tegefaulkes commented Feb 28, 2022

CMCDragonkai commented Feb 28, 2022

tegefaulkes commented Feb 28, 2022 • edited Loading

tegefaulkes commented Mar 1, 2022

tegefaulkes commented Mar 3, 2022

tegefaulkes commented Mar 8, 2022

tegefaulkes commented Mar 8, 2022

CMCDragonkai commented Mar 10, 2022 • edited Loading

tegefaulkes commented Mar 11, 2022

CMCDragonkai commented Mar 11, 2022

CMCDragonkai commented Mar 11, 2022

CMCDragonkai commented Oct 25, 2021 •

edited by tegefaulkes

Loading

scottmmorris commented Oct 29, 2021 •

edited

Loading

CMCDragonkai commented Oct 29, 2021 via email •

edited

Loading

scottmmorris commented Nov 1, 2021 •

edited

Loading

scottmmorris commented Nov 1, 2021 •

edited

Loading

scottmmorris commented Nov 4, 2021 •

edited

Loading

tegefaulkes commented Feb 28, 2022 •

edited

Loading

CMCDragonkai commented Mar 10, 2022 •

edited

Loading