API Slowdowns

[justins]

In the past few weeks, we've been experiencing major slowdowns with API calls, particularly for 360 Field. Has anyone else experienced this?

I also reported an issue to support where it takes 20+ seconds to add a user to a project within HQ through the UI, and they pointed to the API slowness issue as the reason behind it, but said there were no plans to improve it.

Is the slowdown happening across all 360 API calls? Is there a reason behind it?

[mikako_harada]

Hi Justin, 

Do you see slowness happening every API call?  Or could you tell any specific calls?  Is it around adding users?  

You mentioned HQ and API, and we just release preview of Forge API:

cf.  https://fieldofviewblog.wordpress.com/2016/08/02/bim-360-account-level-api-available-as-preview-beta...

So just want to make sure this is not something we are adding unintentionally.

[jsquaredz]

Yes. I reported these same issues. Taking 3 minutes to get a response from the API call. This happens on rather large calls, like all issues, or all equipment, but its only a couple thousand records so it should still be quite fast to generate the 2 meg file. Once generated I get it in 3 seconds, just takes 3+ min to generate.

[mikako_harada · Accepted solution]

For the sake of adding conclusion to this thread - for those who is reading this post and wondering what has happened or is there any reason.  Justin and I had a separate discussion.  

Further looking at it, including the issue with UI, we learned that the slowdown was observed when Glue project was integrated with HQ. The reason is that, before the migration the total number of users are only for Glue users, which is typically a few hundreds; while after the migration, the users are merged with Field users, which is typically thousands. So the number of users for Glue UI to consider for auto complete was drastically increased. So technically speaking, this is not a bug, rather limitation of current approach.  Something that the product team may need to improved in future. This was logged as PDC-15532 in the engineering database.

Retrieving Field issues with API is a separate issue. Above shouldn't be affecting the access to issue data.  And Justin tells us he was able to workaround by retrieving record set by chunk to avoid time out.  

I hope this explains.  

[mikako_harada]

Hi Justin, jsquaredz, and community, 

Long over due, but the enhancement to address this has been implemented and pushed to the production today.  You should see the improvement in the performance. If you are the owner of large team, please give it try!