I am having this issue still with Navisworks alpha. I have tried 2.19 and now 2.20 wip.
anything that is 9k+ objects struggle (even with multiple sends) but sub 9k is seems fine and quite fast.
At the moment I am just sending small groups of objects without issue but the model is very large.
I am using a local docker instance and seems to happen in the final part with the send (74.5% part).
Any advice would be appreciated.
— update —
this seems to be a memory issues with navisworks. with a large number of objects the memory fills and then starts using disk and then runs out of disk and stops. Not sure why so much memory is being used during send (i would have thought if sends objects one by one and then frees up the memory but it seems to keep everything in memory during the send).
Unfortunately I cannot share the model or screen shots… I can describe it a bit and relate some of the issues I have been having.
It is a Navisworks nwd file that is a composite of an number of other nwd and dwg files. the total number of individual objects is over a million (300M source file) and I have been trying to break it down into sub sections. I have had luck with object groups of below 9k but after that it gets tricky (I have had a couple of success with 11-16k but anything too big just gives transport errors for example one was 100k). I don’t really know how many objects here are until I get the the “sending to speckle” stage so it is a lot or trial and error.
I have had one that wouldn’t convert for love or money… it was single object but I suspect it was crazy detailed.
Not sure where the log files are but might be able to share them.
Mostly I get the transport failure and when I attempted a really large file I got the out of memory error. Memory seems to balloon both during conversion (memory error here, or at send) and then during final send (transport errors at this stage).
Ok - I will see if there is anything in the server logs.
The transport errors we are looking into, the memory problem, could potentially be tied to the machine you are using.
Every object selected for publishing will be converted prior to sending and then the sending procedure – because we are sending the entire model hierarchy.
I could look at an option to send individual objects, but this would still be a flat list of objects to convert and then send. However, it might improve memory overhead, though I’m not fully convinced of this.
I am using my own docker server so I will have a look around and see what I can find.
My only idea would be to convert and send for each object individually to save memory. Not sure if you could maintain the hierarchy using detach for the objects that have been sent (via a reference?).
What would be great is to get a count of the objects selected during the prep stage (maybe there is a navisworks option… will check).
thanks again
Tony (ps I am in Australia so my replies are going to be spaced out timewise!)
pps I mean 300M for the source file not G… sorry!!!
