Posts made by atripp

Hi @james-traxler_1560,

Thanks for all of the info!

We are using ProGet Version 5.3.28 (Build 16), and calling the API directly.

Ah, that explains it. We really only test this with the Docker client, which wouldn't trigger this issue.

Usage scenario: removing or renaming tags for a manifest. As there is no specific route for this use case, we are DELETE-ing the manifest through the v2 docker API and re-adding (with PUT) the new/remaining tags.

Is this a Docker API limitation? We were thinking of adding a API for tag management for our own CI/CD platform (BuildMaster), but just worked-around it, and no one ever asked.

we are concurrently executing 6 PUT requests:

This is probably where the issue is. Looking at the database code, there is an opportunity for this to happen between a DELETE and INSERT statement.

It's pretty easy to fix, can you give it a shot? If it works in your test, then I'll commit the change. But note this will be for v6.0.15, so unless you upgrade to that or later, then you won't have the patched code.

https://inedo.myjetbrains.com/youtrack/issue/PG-2140

If you download the file attached to the above link, you can run it against your SQL Server database.

Thanks,
Alana

Hi @Joseph-Saladino_7499 ,

Hmm, that's really weird; the next thing I would try is to just restart all the machines (including the BuildMaster server). That's easy to do and the issue might go away. It could be a software error in the Windows networking stack

It's possible that there's some intermediate network device that's interfering. Router, QoS device, etc. That's harder to diagnose, and would require some tracing ... starting by making sure that the RST packet is actually being received on the BuildMaster server (that's usually what indicates the error message), and then figuring out what device is sending it, and why.

Our code doesn't work at the packet-level, but if the remote server (Inedo agent) is where the RST packets are coming from, it likely has to do with some obscure Windows setting... maybe even firmware on the network card. But at least something to look at.

Cheers,
Alana

Hi @james-traxler_1560 ,

We can help with this...

In general, the error makes sense if you're trying to re-add an image that's already there... but can you describe how you're doing that?

I.e. are you using the Docker client to push an image, or using the API directly? execute the PUT statement?

What is the version of ProGet that you're using?

Can you give a full stack trace (not sure if there's an error)?

Thanks,
Alana

@jramon-ortega_2700 thanks so much for testing this, and verifying the issue is occurring outside of Otter. This is what we were afraid of, but it shows where the problem is.

This is definitely a bug/issue with PowerShell, and likely the way that the DSCR_Application module is designed? We don't really know.

I doubt Microsoft will fix this... but it doesn't hurt to ask. You have a very good reproduction case now.

It might be possible for us to work-around this, but unfortunately it's not a trivial matter. We'd need to find an alternative to "local remoting" to handle process communication.

What would be good to know, how widespread is this?

Is it only this DSCR_Application resource? Maybe the creator of DSCR_Application can work-around the issue?

Using Chocolatey Packages, I guess, could be a good replacement? Maybe there's a way to do it with your own Ensure-based PowerShell script instead of a DSC resource?

We like the ability to use DSC resources, but if it's only one or two, then maybe it's not worth the effort to work-around?

Hi @Joseph-Saladino_7499 ,

The message, "An existing connection was forcibly closed by the remote host" basically just means that the remote server abruptly hung-up the connection.

Based on the fact that there's no errors happening on the remote servers (agents), then I wonder if it's related to firewalls blocking a connection?

It could also be related to the agent refusing to connect because the AES encryption key is not correct. How are the agents configured (with / without encryption key?)?

Hope this at least points in the direction.

Thanks,
Alana

Hi @marc-ledent_9164,

There's really just one, and that's the artifacts (BuildMaster) or packages (ProGet).

To configure server level shared storage, you'll want to configure:

• For BuildMaster, the Artifacts.BasePath value in Administration > Advanced Settings.
• For ProGet, the Storage.PackagesRootPath value in Administration > Advanced Settings.

Do note that ProGet also allows configuration of storage at a feed level, which you can find on the Manage Feed page for each feed. Click change on the right side of the Disk Path property.

This guide might be helpful:

https://docs.inedo.com/docs/installation-high-availability-load-balancing

Hi @jramon-ortega_2700,

I'm not a PowerShell expert myself, but I looked at the code for the operation.

We use the Invoke-DscResource CmdLet to run the configuration...

https://github.com/Inedo/inedox-scripting/blob/master/Scripting/InedoExtension/Operations/PowerShell/Dsc.cs#L140

Not sure if that helps, just wanted to point you in the rightd irection

Cheers,
Alana

Hi @jramon-ortega_2700 ,

Another strange error! I chatted with the engineers here, and can summarize this a bit.

In Otter 2022, we went to .NET6, which has those annoying bugs with modules in a "hosted powershell" environment. So, we switched to using something called "local remoting" - which basically allows us to communicate with the PowerShell process using PowerShell's remoting functionality. So basically, PowerShell is now executing using "PSRemoting" - so it kind of makes sense that a remoting error is happening.

After researching the "serialized xml is nested to deeply" message, this seems to be a powershell bug -- it's an error displaying an error message over remoting, maybe. It happens with PSRemoting and DSC it seems, but who knows

Are you able to try to run this via PSRemoting, to see if you can reproduce the error? Any errors logged in the event log.

A few posts mentioned that this has something to do with security?

If you search "serialized xml is nested too deeply" you should find a handful of things to try... mind giving those a shot, and see if it makes any difference?

Thanks,
Alana

Hi @Rohan-Patel_7666

I think that this issue is related to TLS configuration; basically your operating system doesn't support the TLS configuration required by websites like nuget.org.

Here is a registry work-around that you can try:
https://inedo.com/support/kb/1161/tls-v12-configuration-and-connection-errors

This will force TLS 1.2.

Cheers,
Alana

@jramon-ortega_2700 thanks for the bug-report; this seems to be related another library upgrade, but we'll get it fixed soon - please stay tuned :)

@kichikawa_2913 thanks for letting me know; well, if you see it come again, please do inform us. It may be a caching issue on Linux/containers, but hard to say

@kichikawa_2913 This is really really weird.

Can you try creating a credential again? Perhaps restart the container? I've never seen anything like this... clearly the extension has loaded, since you can view all the components ‍️

Which of the credentials did you create by the way?

Hi @kichikawa_2913,

Do you have multiple web nodes? I'm thinking that, perhaps, one of the web nodes didn't load the extension?

Hw about if you lick on TeamCity in that list? It should enumerate all of the components in the extension.

You were able to create it... so we know it loaded at some point...

Alana

@kichikawa_2913 we don't have a typical set-up per se, really

A lot of set-ups are very simple, since the team is new to CI/CD, and they just have the BuildMaster server do everything.

Others have dozens of build agents configured in a pool to service various teams, and use lots of shared configuration.

Most are somewhere in the middle -- since they started simple, and gradually explored over time. That's generally what i'd start with -- tackle one pattern/team, then move on to the next and improve over time.

Hi @kichikawa_2913 ,

Given that you're already familiar and comfortable with using build agents, then we'd recommend a similar set-up with BuildMaster. To do this, you'd provision/configure your build server(s) and then have BuildMaster connect to them, generally using the Inedo Agent.

One common pattern is to create a role, such as build-server, and then target it in your pipeline stage, like this:

This way, it will act as a server pool. You can also do this in OtterScript as well, for some very fine-grained control; it's considered a resource pool.

In any case, whatever server you target is where the code will retreived from source, built, and tested.

hope this helps!

Hi @gurdip-sira_1271 ,

We don't think anything changed with regards to Git Rafts & Otter v2022, and don't have any issues on our end with them.

Is this a new instance with them or something?

Cheers,
Alana

Hi @jramon-ortega_2700 ,

This should be fixed now! If you go to Admin > Extensions, can you download Scripting 2.0.1?

Thanks,
Alana

Hi @jramon-ortega_2700 ,

We just discovered this issue with Otter v2022; it has to do with a bug in the hosted PowerShell runtime, .NET6, and loading certain modules

It doesn't look like Microsoft will fix it unfortunately, but we're currently scrambling to work-around the issue and get it fixed ASAP.

Cheers,
Alana

Hi @zakaria-smahi_2127,

That badge looks like something generated by a CI-server, and I think it's showing the latest version or tag or something related to a release branch? I don't have much context.

It's not something that ProGet (a package manager) would generate, since all of the packages in ProGet are considered "published".

Our CI/CD platform, BuildMaster, can generate these badges:
https://docs.inedo.com/docs/buildmaster-ci-cd-continuous-integration-server-badges

Cheers,
Alana

@gurdip-sira_1271 it's possible there's a bug in the visual mode with that, especially w/ rafts...

Here is how you you would call scripts from another raft:
PSCall RaftName::Set-WebServer;

When you don't specify a raft name, then Default is used as the raft.

Hi @gurdip-sira_1271 ,

If your script has no parameters, then there's nothing to really show on the Visual Editor. However, you should be able to save the statement.

Behind the scenes, the visual editor should just make a PSCall Set-WebServer; in text mode. What does the text mode look like?

Cheers,
Alana

Hi @gurdip-sira_1271,

I guess the bit I am confused about is if a restart is needed mid script. Would the agent on the server resume the script post reboot?

You would need to perform the restart in OtterScript.

for server REMOTEMACHINE
{
   PSCall MyScript1.ps1;

   Restart-Server;

   PSCall MyScript2.ps1;
}

For script assets, how would I reference from my Powershell script? Or should I use OtterScript and then have the step above and then my script?

In this case, you'll want to use OTterScript to deploy the script assets first.

To add, I get this error when making a job template:

That's a regression, and we'll fix it pretty soon. It has to do with Job Template folders in Git Rafts. We're shipping Otter v2022 today actually :)

In addition, when I drag a ps1 script in an otterscript orchestration, the popup is blank and i can't add the script.

Does your script have any parameters? If so, that is to be expected. Its should save if you hit the save button however. Note that in Otter v2022, shipping very soon, the script editor is totally rewritten and should be fixed.

Cheers,
Alana

Hi @gurdip-sira_1271

There is a Restart-Server OtterScript operation that you can use. You can specify a timeout/timer.

You'll want to store that chocoPackages.txt file as a script asset, and then deploy it to the server you're running the scripts on:

Get-Asset chocoPackages.txt
(
    Type: File
);

However, you may want to consider this approach for your chocopackges:

https://docs.inedo.com/docs/otter-howto-ensure-chocolatey-packages

Hi @gurdip-sira_1271 ,

I want to share this guidebook that we wrote;
https://inedo.com/support/resources/ebooks/windows-first-iac-cca

Basically it discusses and walks you through how to CI/CD with your "infrastructure as code" using Otter and BuildMaster (our CI/CD tool). The screenshots are outdated, but after a lot of research, our conclusion is that the "pipeline" approach is likely the best.

Cheers,
Alana

Hi @chris_5235,

Unfortunately I'm not really sure what's happening, but I suspect it has to do with an HTTPS redirect?? I would look there next.

As you mentioned in the OP, the package can be download from the endpoint, so long as you provide the credentials. So, I guess, PowerShell isn't sending credentials after a redirect, perhaps?

One thing that you might want to do, is use a tool like Fiddler to capture the request/responses. That will let you verify that the issue is indeed PowerShell not passing credentials to ProGet. You can look at the authentication header.

Things to look at:

• make sure ProGet is sending an https download link for the package file in the FindPackagesById() request
• X-Forwarded-Proto is set to be https
• If you must, set BaseUrl under advanced settings

Cheers,
Alana

Hi @mcascone ,

I think that was a feature that got lost in 5.3, with the new UI.

Easy enough to bring back, I think, so we'll add it via PG-2134

Cheers,
Alana

Hi @gurdip-sira_1271 ,

Your Git repository must be structured in a particular way to work as a Git Raft; but it's best to let Otter manage all of that. Here is a sample: https://github.com/Inedo/OtterRaftTest

The repository you use for a Git-based raft shouldn't be used for anything else - just raft storage.

Cheers,
Alan

Hi @gurdip-sira_1271 ,

Great, sounds like its working :)

I think the issue is that you didn't have the folder-structure in place. Scripts should go in the /scripts folder basically.

Cheers,
Alana

@nmarkkula_7220 make sure the v3 API is enabled on your feed; the endpoint will look like https://MY.FQDN/nuget/MYNUGETFEED/index.json

visual studios detects the API once it queries it, so if you changet the URL then it should work

@nmarkkula_7220 I'm not totally sure, but I'm pretty sure Visual Studio used to pay attention to that flag. Perhaps it's a regression or something in Visual Studio?

But it looks like that's on the v2 API, and Microsoft (and we) aren't really maintaining that anymore. I would try using the v3 endpoint instead.

@nmarkkula_7220 this can be a confusing issue

Listed/Unlisted is a serverside metadata, like the download count. When you cache or pull a package from nuget.org, then ProGet won't have severside metadata. This is why download counts start at zero once you upload/pull a package to ProGet.

You will have to unlist the package in Proget too.

Hi @luke_1024 ,

Unfortunately this just doesn't look like something that's going to be easy/trivial to fix. Maybe it's a bug in the compression library we're using, or maybe cargo-deb is doing something unusual with the compression format that we're not expecting, or isn't documented in the normal debian commands.

It looks like you recreate the package with dpkg-deb then it works.

We will upgrade the libraries in the next major version, and maybe it will go away - but for now it doesn't seem worthwhile fixing, To date, just a single user (you ) has reported this, and if we hear from customers, we'll consider it too.

In the meantime, the workaround of repackaging should do the trick.

Hi @mcascone,

To do this, you'd want to download the latest version of the package, edit it in a zip file, then reupload it. A little tedios but fastest way I think

Cheers,
Alana

Thanks for letting us know, @hwittenborn

We must not have the http --> redirect setup in our DR server , we'll try to get that soon

Hi @rmusick_7875

The sort-order for feeds is not configurable the ProGet UI, but some tools (like Visual Studio) do allow for sorting.

Otherwise, we haven't had any other requests for improving the ProGet UI for Chocolatey Feeds (especially not from customers / paid users that we're aware of), but if we hear more requests we'll definitely consider improvements like this.

Hi @alin-kovacs_4228 ,

We don't maintain documentation for that version I'm afraid; the Native API is available, and you can find it in the software with /reference/api

However, your best bet would be to work with our team to figure out what you need to do and get some help migrating to a newer version or something :)

Cheers,
Alana

Hi @marc-ledent_9164,

The high-availability / cluster configuration can be a little tricky... but glad that changing it worked.

The message "More than half of the servers are in an error state, consider restarting the BuildMaster service." must be a cached error message? We try to detect if there's a major problem with the service / agents, and then trigger that message.

Can you try to restart the BuildMaster service on each of the nodes, and see if it goes away?

Cheers,
Alana

Hi @rmusick_7875 ,

No problem, happy to help.

[1] I'm afraid it's not really possible to sort the packages; they're displayed in the order that's returned from the remote feed (i.e. chocolatey.org), and sorting isn't possible when requesting results. Typically it's done by popularity, but sometimes they come back with recently-updated. I'm afraid it's not predictable. Internally, ProGet sorts a feed by recently updated.

[2] We used to do this, but it creates a lot of confusion because many package titles look like they are a package ID. It was especially confusing for packages with a title like Initech.Utils, but with a ID of InitechUtil2 or something. NuGet.org also changed similarly.

We may consider changing the behavior for Chocolatey feeds.

Hi @marc-ledent_9164 , sorry on the slow reply, I wasn't so familiar with OpenShift so I wanted to research a little.

First, I think your Service.MessengerEndpoint should be tcp://*:4242, because you don't know which node will be active. It might be buildmaster-0, but it might not.

What also isn't clear, do you need to "open" or otherwise map port 4242? I'm thinking the service messager is working on the node that can connect to itself, but the nodes aren't communicating over the internal network.

Cheers,
Alana

Hi @marc-ledent_9164 ,

BuildMaster doesn't use a reports directory... I'm guessing someone configured that to store reports generated from custom queries using something like SQL REporting or something? Or maybe PDFS/screenshots for auditing?

Cheers,
Alana

Hello @bkohler_1524,

Because packages are immutable, you need to decide the future-compatibility at the time you create the package. This is where Semantic Versioning dependencies can really help.

Following the rules of SemVer, you can say that "ProductA 1.0.0 has a dependency on ProductB (1.1.0-2.0.0]", what that means is that you can use every version of ProductB from 1.1.10 up to (but not including) 2.0.0.

This allows you to make as many minor versions of ProductB (such as 1.200.0) before "breaking" compatibility.

If these version numbers are business/marketing-driven, and it's impossible to change their mind on versioning, you can always maintain an internal version number - this is what many products (such as Microsoft) does.

Hi @jon-benson ,

We don't have a tutorial for that I'm afraid :(

However, you're on the right track -- it does involve using the Advanced Properties on the User Directory to specify the domain controller host, credentials, and so on.

Since you mentioned you have / will purchased ProGet, let me reach out to my colleague Gene, he is our Customer Advocacy Manager and often sets-up appointments to help with onboarding/configuration for new users. This could be a chance to get some assistance on that.

Thanks,
Alana

Hi @jon-benson

Sorry for the mix-up / confusion. I'll try to clarify a few points.

LDAP/AD integration is only available on paid editions of ProGet (Basic, Enterprise), and you can integrate with Azure AD using LDAP. This requires you to type in your Azure AD username/password in the ProGet login page.

SAML/Single-Sing-On is only available in ProGet Enterprise edition. This allows you to sign-in to ProGet without typing in your Azure AD username/password. Also, this is no longer a preview/beta feature (the docs were just incorrect; I've now updated).

For ProGet v2022 (aka 6.1), we are developing an improved Security Management user interface. It's basically the same, but just easier to work with. It's available as a preview feature, and that preview is enabled by default on new installations. We just added SAML support to this new interface, so on 6.0.12 you should be able to use it.

hope this helps clarify,

Alana

Hi @nuno-ildefonso_8876 ,

Thanks for that information; this will be fixed in the next maintenance release (BM-3788), later this week (on Friday).

Hi @pariv_0352 ,

Are you able to see the results of the "DockerGarbageCollection" job?

This is actually what's responsible for deleting those images, and it runs nightly by default.

Let me share the code to it; if you can understand the database structure already, then hopefully it will help you to identify why it's not workikng, and what you might be able to look at in the logs to help troubleshoot:

[ScheduledTaskProperties(
ScheduledTaskTypes.DockerGarbageCollection,
"Deletes unreferenced Docker blobs.")]
public sealed class DockerGarbageCollectionTask : ScheduledTaskBase
{
    public override Task ExecuteAsync(ScheduledTaskContext context) => this.GarbageCollectAsync();

    private async Task GarbageCollectAsync()
    {
        using var db = new DB.Context();

        this.PercentComplete = 0;

        var usedBlobs = new HashSet<DockerDigest>();

        this.LogDebug("Gathering list of all rooted blobs...");

        foreach (var image in db.DockerImages_GetImages(Feed_Id: null))
        {
            if (image.ContainerConfigBlob_Id.HasValue)
                usedBlobs.Add(image.ContainerConfigBlobDigest);

            DockerManifest manifest;
            try
            {
                manifest = new DockerManifest(image.ManifestJson_Bytes);
            }
            catch (Exception ex)
            {
                this.LogError($"Image {image.Image_Digest} has invalid manifest: {ex.Message}");
                continue;
            }

            // "Fat images" do not have blobs as layers.
            if (manifest.Layers == null)
                continue;

            foreach (var l in manifest.Layers)
                usedBlobs.Add(l.Digest);
        }

        this.LogDebug($"Found total of {usedBlobs.Count} rooted blobs; finding unreferenced blobs...");

        var allBlobs = (await db.DockerBlobs_GetBlobsAsync(Feed_Id: null))
            .Where(b => b.Feed_Id == null)
            .Select(b => DockerDigest.Parse(b.Blob_Digest));

        var unreferencedBlobs = allBlobs
            .Where(d => !usedBlobs.Contains(d))
            .ToList();

        this.LogDebug($"Found {unreferencedBlobs.Count} unreferenced blobs.");

        using var fileSystem = new DirectoryFileSystem(ProGetConfig.Storage.DockerBlobStorageLibrary);

        for (int i = 0; i < unreferencedBlobs.Count; i++)
        {
            this.PercentComplete = (i + 1) * 100 / unreferencedBlobs.Count;

            var digest = unreferencedBlobs[i];
            this.LogInformation($"Deleting blob {digest}...");
            if (!ProGetConfig.Feeds.RetentionDryRun)
            {
                await db.DockerBlobs_DeleteBlobAsync(Feed_Id: null, Blob_Digest: digest.ToString());
                await fileSystem.DeleteDockerBlobAsync(digest);
            }
        }
    }
}

Hi @NUt ,

LDAP / Active Directory integration is a paid feature, and should not be available in ProGet Free Edition. We don't intended to change that in our next version (v2022).

If you're able to log-in or use it, then it's most certainly a bug of the new Security Management preview features. Please don't rely on that, because it will probably not work in a newer version :)

Cheers,
Alana

@mcascone I don't think it's new, but it's just used to specify the dependencies field in the manifest file; I'm thinking, perhaps, it might be similar/identical to the consumes field you added?

https://docs.inedo.com/docs/upack-universal-packages-manifest

The only thing Inedo tools use it for today is just displaying information in ProGet, on the dependencies tab. That may / may not be helpful.

@ales-bahnik_2824 thanks; so that tells us the call is making it to ProGet OK, and there are no permissions issues. If there were permissions problems, you'd get an error like you did above.

Not sure what to do from here. I tried very quickly to reproduce, and the package deletes fine

1. New feed, connect to NuGet.org
2. Download NewtonSoft.JSON 3.0.1, then verify package is no longer remote package in UI
3. Run below powershell, then verify package is now a remote package in UI

PS C:\Users\atripp> Invoke-RestMethod -Method Delete -Uri "http://proget.localhost/nuget/nugets/package/Newtonsoft.Json/13.0.1" -Headers @{"X-NuGet-ApiKey"="058fca3993cde88d771b142b876913a0a126f16b"}

The delete will not produce an error if the package doesn't exist. So maybe it's already deleted.

Otherwise I'm not sure how else to debug. The method the API and Web page call are exact the same.

hi @ales-bahnik_2824 , what happens if you provide an incorrect API key? Do you get an error message?