Troubleshooting Error Page in Traces for Version 7.12.2

gangaprasad

·Feb 17, 2025 11:22 AM

Hi, I was getting this error page when trying to load traces. I was using version 7.12.2 on a self-hosted setup.

Arize News

Troubleshooting Error Page in Traces for Version 7.12.2

gangaprasad

·Feb 17, 2025 11:22 AM

Hi, I was getting this error page when trying to load traces. I was using version 7.12.2 on a self-hosted setup.

56 comments

· Sorted by Oldest

gangaprasad
·
Hitting the memory limit—what were the memory recommendations for Phoenix?
Ben M.
·
How much memory do you use today? Ours (Also self hosted) is running OK with 1gb of memory. We’ll probably double that in the near future, but we don’t get this error
gangaprasad
·
i kept 4Gb limit
Anthony P.
·
Hi there, thanks for reaching out! Have you been able to view the trace page previously or have you always seen this error?
gangaprasad
·
before i'm able to see
Anthony P.
·
Did you perform an upgrade or change any config before seeing the error?
gangaprasad
·
no after seeing this error updated to latest
Anthony P.
·
Can you share the URL you are trying to visit? Just the path is okay if you don't want to share the whole thing
gangaprasad
·
this is the path /projects/[project_id]
gangaprasad
·
this is the project
Anthony P.
·
Thanks for the info! Now about the memory limit, are you seeing a warning or error about a memory limit?
gangaprasad
·
it was crashing
Anthony P.
·
So when we load the trace/span/session view we load the contents of the spans into memory before serving them over the API. Typically spans are much smaller and we can paginate but in this case your spans are quite large. For now I think we can do two things:
If you increase the memory limits on your container, it should stop crashing for a while
I have filed an issue so that we can try to reduce memory usage on those pages for the case when spans are extremely large https://github.com/Arize-ai/phoenix/issues/6417
Anthony P.
·
Sorry for the inconvenience! Super interested in hearing about your use case that generates spans with so many tokens 🙂
Tharun P.
·
So these are traces for agent runs with hundreds of documents with text and metadata. I am able to view them fine on langsmith though