Maybe I'm missing something here but the error indicates that cmdportal.com.Controllers.ContentController which doesn't have a parameterless constructor and has nothing to do with NuCache. How does this error lead back to the cache you think?

FYI: you're unable to disable NuCache.

Hi Sebastiaan,

The ContentController is inherited from UmbracoApiController and that one has a parameterless constructor as I remember.

Also this controller works fine when we call it manually or one by one, it is just that the cache starts to throw errors at some point.

In my understanding it tries to do a snapshot but something goes wrong, probably because of the multiple threads.

...
at Umbraco.Web.PublishedCache.NuCache.PublishedSnapshotService.GetElements(Boolean previewDefault)
at Umbraco.Web.PublishedCache.NuCache.PublishedSnapshot.get_Elements()
at Umbraco.Web.PublishedCache.NuCache.PublishedSnapshot.get_Members()
...

As it is working half of the time, I think the issue is not with the missing constructor.

Half of the times the call does not reach the controllers method, the client just gets the error message.

We also receive errors under Umbraco when we try to access the back office while the migration is running that also relates to NuCache. Unfortunately I don't have that error message. As I remember it has a weird "panic" message.

Right.

I think it might just have worked by fluke in v7, and I bet you had some weird caching errors there.

I think we'll need some sample code to see what you're trying to do.

I have copied the article import code below. (stripped out some irrelevant lines to make it short...)

   [RoutePrefix("umbraco/api/content")]
   [BasicAuthentication]
   public class ContentController : UmbracoApiController
   {

    [HttpPost]
    public IHttpActionResult ImportArticle([FromBody]object objarticle)
    {
        try
        {
            CommonClass.BlogArticle article = Newtonsoft.Json.JsonConvert.DeserializeObject<CommonClass.BlogArticle>(objarticle.ToString());

            IContentService cs = Services.ContentService;

            var content = cs.CreateAndSave(article.title, this.folderID, "article", 0);

            content.CreateDate = article.datecreated;
            content.UpdateDate = article.lastEdited;

            //content tab
            if (sharedData.pathchanges != null)
            {
                article.maincontent = ReplacePathsinmainContent(article.maincontent);
            }

            content.SetValue("main", article.maincontent);
            content.SetValue("publishName", article.publishName);
            content.SetValue("priority", (int)0);

            var result = cs.SaveAndPublish(content);

            if (!result.Success)
            {
                Logger.Warn(this.GetType(), "Failed to add article " + content.Name);
            }
            else
            {
                Logger.Debug(this.GetType(), "Added the article " + content.Name);
            }

            return Ok();
        }
        catch (Exception exc)
        {
            Logger.Error(this.GetType(), exc, "ImportArticles");
            return this.InternalServerError(exc);
        }
    }
   }

At the moment we are working on to replace the CreateAndSave with CreateContent, could that help?

Things to note:

The error "An error occurred when trying to create a controller of type 'ContentController'. Make sure that the controller has a parameterless public constructor." does not indicate that the controller does not have a parameterless, public constructor, but that something went wrong when constructing the controller. Here: getting a snapshot from NuCache fails.

Then, although NuCache has been tested under severe load... it may be that you are triggering an issue that only occurs under heavy, multithreaded load. You write: "As I remember it has a weird 'panic' message." and well, these 'panic' messages happen in NuCache when a situation that just is impossible... happened nevertheless.

So... there definitively is an issue behind all this. I am interested in anything you can give me - stacktraces, details about the panic messages, description of how your import process works, etc - that would help me better understand the problem + create a proper issue on our tracker.

Stephan

Looked deeper into the exception that you posted and hey, look, the panic error is there:

    "Message":"An error has occurred.",
    "ExceptionMessage":"panic",
    "ExceptionType":"System.Exception",
    "StackTrace":"   at Umbraco.Web.PublishedCache.NuCache.ContentStore.CreateSnapshot()
        at Umbraco.Web.PublishedCache.NuCache.PublishedSnapshotService.GetElements(Boolean previewDefault)
        at Umbraco.Web.PublishedCache.NuCache.PublishedSnapshot.get_Elements()
        at Umbraco.Web.PublishedCache.NuCache.PublishedSnapshot.get_Members()
        at Umbraco.Web.Runtime.WebRuntimeComposer.<>c.<Compose>b__0_2(IFactory factory)

There is only one place where a 'panic' can occure in the CreateSnapshot method. And... we have already seen one case where this exception occured. So, this confirms that we do have a multi-threading issue with NuCache.

Going to link to the GitHub issue here in a few.

Side note: your code is OK and there is no "magic" (replacing CreateAndSave with CreateContent etc) that I am aware of that would work around the situation.

Hi Stephen,

happy to help :)

See my sample code above. We are calling it from another website with RestSharp. There are more than 30,000 contents to create (in v7 it took around 8 hours). To speed it up we are creating multiple Tasks and do it in batches. At the moment we use 500, the same number of tasks as with v7. Then wait for them to finish and then start another 500.

Let me know what else could I provide and I will try my best.

At the moment we are working on to replace the CreateAndSave with CreateContent and check if that helps.

We were looking at that because we noticed in the cache table the draft is saved into it and also the published saved into it too.

is working on a reply...