yea but that breaks even more things, pretty much cant login then 
anyhow its not a UID issue, or a user issue…. and never was
getting closer to it..
flame01 → still has errors
flame02 → no errors
always “link” issues, so I stack traced it to flame trying to hardlink to/from appleDouble files *_ which fails !!! I had something similar before where it would corrupt projects by filling them with these shitty files.
ok so flame02 , opens project, saves something → no issues and no AppleDouble files created on the NFS mounts 
flame01 → DOES create a bunch of AppleDouble files!!!
replayboys@rpsresolve001 test8889 % cd /opt/mnt/flameprojects26/test8889
find . -name '._*' -type f -print
./catalog/Workspace.wksp/._.workspace.000.wks
./.volatile_6490_501_20@1527B37B-01CB-4DCB-8954-A98F9A764ED8/._h01195a7f_6946ae82_000e19bf
./.volatile_6490_501_20@1527B37B-01CB-4DCB-8954-A98F9A764ED8/h01195a7f_6946ae82_000e19bf/._pure_side_car_setup.resize_node
aha? smoking gun? maybe?
I looked once for a way of blocking them. They also create mayhem on Avid Nexis.
Don’t think there’s a way to tell Apple to keep that crap to themselves. But could run a cron jon that keeps nuking them.
yea the problem is more that one machines creates them, and yet another does not,
the one that creates them cant save timeline resizes..
i can see them beign created live, but just on that one pesky box, and i have no clue why. and also why they cause havok. BUT only on one machine, the other just ignores it, same macOS, same everything… ughhh
I might be reaching here, but some apps that I have used have had file requesters that would “test” every mount point before they would let you save something, even if the file requester dialog was set to a valid device. Nuke had this bug for some time. If, somewhere in your “recently saved files” list, there was an old mount point that was no longer connected, the file dialog would pause for several second per each missing device it scanned, timeout and then save the files. Took me forever to track down. Caused a lot of ‘mystery pauses’ at seemingly non-related times.
Is it possible that you have an old mountpoint that is somehow being checked, timing out, or trying to check it’s connection and failing for a device that is no longer present….even though you aren’t trying to save to that device?
on ZFS there is an attribute to make the storing of xattrs (which is what the Apple Double files are essentially) more optimal. Not sure how Qnap deals with that, but the proper tunable is:
zfs set xattr=sa $POOLNAME
Although it makes no sense why there is a discrepancy between two machines, at least setting the above it also a best practices.
also this ChatGPT page has great info regarding this:
I am getting the same on a xfs backend that isnt qnap but ill mess with this thanks
yea me and chatty have been trying to stacktace this bullshit together for the last 80hours hahah
the weird pause is only when I use /hosts// as my project home , even on a brand new proejctserver AND brand new linux flame client , fully snadboxed
I think yhe /hosts/… stuff is somehwat special , like it sends the data to somewhere and then project server/host does something with it? no clue.
this is pretty interesting info:
Check to see if the NFS is being mounted differently between the working and non-working machine.
For instance is one machine mounting as NFSv4 and the other NFSv3?
NFSv2 and NFSv3 do not support these operations over the
wire. The getting and setting of ACLs are not supported but extended
attributes and resource forks for MacOS clients are supported by storing
extended attributes and resource forks in an "Apple Double" (also known
as a dot under bar file) files on the server. On the other hand NFSv4 can
optionally support ACLs. And can optionally support the storing of
resource forks and extended attributes as NFSv4 named attributes.
I went full crazy and compared every background service, every unning process , every nfs, smb, whatever defaults anything .
Machines are identical. there is only one thing with the working machine thats different that I cant talk about - i will have access to more machines in a few days (people are still going hard) to maybe figure it out .
We never specifically mount or access anything via /hosts. We always have a predefined NFS mount for our NAS.
yea just observing that saving to /hosts (which is default for flame when connecting to a projectsever when nothing is in sysconfig!!)
is very slow but doesnt give me link errors.
Saving to a specific mountpoint , even if its the SAME nfs share mounted with the same options → fast but errors on the one box.
i really cant say why, how would the process of writing to /hosts/projects be different than /opt/mnt/projects… if its the same nfs export beign mounted in 2 different places. (this is the same for linux client) even in a complete sandbox writing to /hosts is god aweful slow
crazy huh?
All link errors are actually related to flame (or rather DLMPD) trying to link around apple doublefiles
I recently had to setup a full sandbox mirror of our infrastructure for ADSK MTL to teradici into for 2 weeks to resolve a deal killer 2026+ bug. A full mirror. NAS, FreeIPA, Project Server, Workstation, Burn. I made a list of basically EVERY command I used to configure this stuff. I provided it to ADSK so they could also review to see if there was anything weird I was doing. It was all legit. They eventually “fixed” the bug, but in the end they didn’t really fix it, but just made it so Flame doesn’t crash. Anyway, I’ll DM you this list, maybe it has something that might help. Still not convinced there isn’t a strange configuration problem on your end.
do you know if the mount point in /hosts you are writing to is going over 1gig or 10gig?
10G, its across 2 different VMs as host and one phyiscal baremetal machine, even running fio on the /hosts mount is fine and shows good speeds.
If you can - and dont have it all closed down - if you can try this on your end i would really appreciate it, start flame with empty sysconfig → select a project server → create a new project leave all on default (so it uses the /hosts thing) → open batch, make a node, save the batch.
this save takes 5s for me, even in a utter simple sandbox between a linux client and a project server, with nothing else
The same is true for me on 2025 project server btw, saving to it takes the sane amount of time
maybe for /hosts flame is doing something special
I think it would be helpful for you to compile a list of EVERY command you are using to setup ALL machines within your sandbox. After all this, I just don’t see any other way of proceeding in helping diagnose this issue. Literally, everything, starting from scratch, for EVERY system involved. Otherwise, we are always going to be missing some part of the picture. And not just a summary, but verbose.
This code can sometimes be more useful than history
# Ensure the log directory exists (create it if missing)
mkdir -p /var/tmp/logs && \
# Generate a timestamp string like 20251220_084700
ts=$(date +"%Y%m%d_%H%M%S") && \
# Run your command, capturing stdout and stderr separately
some_command \
> >(tee -a /var/tmp/logs/${ts}-stdout.log) \ # Redirect stdout into tee, append to timestamped stdout log, still show on terminal
2> >(tee -a /var/tmp/logs/${ts}-stderr.log >&2) # Redirect stderr into tee, append to timestamped stderr log, still show on terminal
tmux is a great way to simultaneously execute commands on multiple machines.
Had the weekend now to test ALL machine sin the office and not just my sandbox with 2026
→ All the M2 Ultra and M4 pro macs and linux boxes have no issues (apart from slowness when using /hosts) , qnap is fine, rocky nas is fine, VM is fine, no errors, stuff gets saved and is fast
→ My M1 ultra test box i had set for the sandbox test is the only mac going crazy with a bunch of link errors
As soon as the M1 touches any project , all other machines show errors in the log about not beign able to parse files etc, so i was put on a wild goose chase thinking they all fail but its really just the one machine aggresively writing appledouble files everywhere, the others dont do that.
took me forever to find that out.. multiple users on the mac, AD joined, manual user that maches the UID of the server etc, all sorts of random link errors and appledouble files trying to be hardlinked etc.
Still trying to figure out what causes the 1 mac to go nuts, I uninstalled everything apart from flame - still going crazy , resetting this thing fully next tomorrow - even observed weird errors when running local projects where its trying to access lock files that dont exist - great! (not) .