It's costing us a ludicrous amount of money

we did not receive a spike in traffic, in fact our function invocation numbers are normal

it's just the duration that's bad

there were no code pushes or deployments at the time this spike started

this seems uncorrelated to any specific factor

and everything seems to have gone up indiscriminately

not correlated to:
status code
user agent
http referrer (where user came from)
IP
request path
none of these categories have values that are disproportionately high usage


in fact, in each of these categories, every single value has gone up. eg if i group by request path, every single request path has gone up significantly in function duration usage

so none of these factors are affecting the spike

minimum function duration spiked, along with 50th percentile, average, 75th percentile etc

it's not one type of request eating all the function duration. EVERYTHING increased significantly

was the project updated somewhere in that time? -> could be some faulty code

I would suggest you to contact support

We did, hoping for a response soon

No, there was no code deploy at that time this could be linked to

we did deploy code AFTER the spike started (before we found out about the spike) but i don't think those two are correlated

also after the spike it's relatively stable/flat (so was stable through that code deploy)

just an insane spike

Since we're not a Vercel server we can't help with any of that at all. You'll have to wait for Vercel support to get in touch with you and in the worst case shut down your deployments to mitigate this temporarily.

appreciate it, hoping for a response from them