We have a Huggingface Space (nllg/DeTikZify) that has worked fine in the past with our builds usually taking around 45 minutes. However, our most recent factory rebuild has been hit with a ‘build error - job timeout’ error after 30 minutes. Is this timeout a new restriction imposed on spaces? Is it possible to increase the timeout to the previous value? We tried setting startup_duration_timeout to a higher value but it did not have any effect. We are using a docker space. Any help would be greatly appreciated.
1 Like
The documentation does not appear to have changed, so this is probably not an intentional change in specifications. There may be a problem with the site, so I will ping it just in case. @meganariley @pierric
startup_duration_timeout: string
Set a custom startup duration timeout for your Space. This is the maximum time your Space is allowed to start before it times out and is flagged as unhealthy. Defaults to 30 minutes, but any valid duration (like1h,30m) is acceptable.
Hi @potamides We saw your email to support and appreciate you posting here for the community as well. 
We’re looking into this and I’ll update you soon.
2 Likes
Thank you!
Thank you all for your support. As of now we have worked around the issue by using a base image that we created locally which does all of the heavy lifting. But we would still be curious to know what the root cause of this issue is since the way our docker spaces are being built has clearly changed. For example, another issue we ran into with our workaround is that we can no longer preload models in our Dockerfile (similar to the preload_from_hub option for normal spaces) since attempting to do so now leads to OOM errors. This has worked without issues for over a year.
1 Like
That’s interesting.
When a function is not malfunctioning but has stopped working altogether, it is sometimes suspected that there is a problem or that it was forgotten to be turned on after an upgrade.