I exported a project (~55 MB) and imported it again on the same instance (this was for testing the export/import feature on this instance). The import showed an error message (504 gateway timeout) but seemed to work (the project is visible). However, now I don’t have WRITE access in the project, despite being Domain Admin.
Is that intended, or did something go wrong here? If it is intended, how can I get access? By explicitly giving myself access via project settings?
Hmm, just when I finished writing this, I got access. It seems like it takes some time (in my case ~12 minutes) for some postprocessing until the project is fully ready. Can somebody confirm this behaviour? Maybe @peter.dahlberg ?
It’s possible, we lock the project until everything is imported, esp. the content of Data Tables and Models. So there is a time frame during the import process where the project exists but it not yet usable.
The gateway timeout may or may not be solvable by re configuring the reverse proxy but it should not matter that much. As long as the zip file was completely uploaded the import runs by itself, you just miss out on any error/warnings generated.
Why 55Mb take that long is another question. 
Great, thanks. It probably takes that long because the instance is really slow in general (probably because the underlying filesystem is running on HDDs).
Your problem description reads very similar to an issue we had with another project.
Due to a bug in the workflow processing, the import of projects will be unexpectedly slow if it contains many workflows - the bug is fixed but not released yet: https://jira.intranet.onelogic.de/browse/QA-2685
Last but not least, we already have a technical refinement for addressing the timeout and false error message: https://jira.intranet.onelogic.de/projects/TR/issues/TR-450
Followup question @adrian.berndl @peter.dahlberg: Do you have idea how to estimate the time it takes to import a large project (~100 WFs, ~50 Data Tables and 56 GB in size)? The project was imported old-style (copy and extract ZIP onto the server in the projectImport folder, send request to /projects/import).
Does it mostly depend on the speed to copy/move the parquet files to the correct location? Does it depend on the database (to create the metadata records for data tables)?
Given there are no longer any severe inefficiencies importing the metadata i would guess it’s transferring the parquet files most of the time.
To also answer my own question a bit with some values from experience: A project with 225 GB size (non-zipped, mostly Data Tables) took ~57 minutes to import on a cloud instance with premium storage.