New group creation error & no Homepage blocks

[marianthi-design]

Hi @tim-hunt
I've just created a group called "NHS Patient Safety" and I received an error.

Although I received the error, the group was indeed created but didn't have the standard blocks on the homepage.

This has happened in the past when CDS would release something.

[tim-hunt]

The issue was the examine indexes had become corrupt and so when trying to create a group it was attempting to update the index and failing, it is linked to this issue: https://our.umbraco.com/forum/umbraco-7/using-umbraco-7/61711-Failed-to-retrieve-indexer-details-

After investigating it appeared to only be an issue on the current production slot and not staging. Deleting and rebuilding the index on the production slot fixed the issue and groups successfully created the new page.

There is a chance we will see this issue in the future and if so I think on startup rebuilding all of the indexes could be a simple but effective solution. For now I'm not doing this but see here for how to do it should we need it https://www.jondjones.com/learn-umbraco-cms/umbraco-7-tutorials/umbraco-search/how-to-rebuild-your-umbraco-examine-indexes/

[simplybenuk]

This looks to have happened again. If this happens the workaround is to manually delete the index and restart the app.
We will look at implementing a proper fix for this.

[tim-hunt]

It appears it was not a one off but a recurring issue. it apepars to be linked to the known issue with Umbraco in Azure with the lucene indexer umbraco/Umbraco-CMS#8006 (comment). The issue is not resolved still and so I have followed the suggested work arounds,

Updated terraform to set the following:

WEBSITE_DISABLE_OVERLAPPED_RECYCLING = 1 - basically stops file locking issues on slot swapping, I doubt its the cuase but its on the list of required settings

In AppSettings:

Update Gloabl:MainDomLock to SqlMainDomLock from FileSystemDomLock - this should fhopefully fix it as it removes the file system as the dom lock.

Changes are pushed to prod so lets keep this open for a few days and try creating a new group on Friday and Monday to see if we encounter the issue again