Harbor heavy load

[sashalarsoo]

Hi harbor team.
We deployed a proxy cache working well with default values.
We tried the proxy for like 9 nodes pulling a 1.5gb image working well.
Now we tried with 100 nodes pulling the image (its mandatory for us for each node to pullthis image) and all the nodes are like in imagepullbackoff despite a low cpu usage by harbor registry in our monitoring. Is there a way to fix this ?

[wy65701436]

Let me clarify: You have 100 nodes pulling the same images, and even you're noticing low CPU usage on the Harbor registry, but still seeing some pull requests fail, correct?

Could you provide more details, such as the core logs, CPU and memory usage of Harbor core and the registry, as well as the database connection status?