Frequent oom

[baiyibing123]

vector performance is very high, I process nearly 200w logs per minute from kafka, only need two pods, each pod only needs 200m memory, but when my vector has a problem, the log delay after 10 minutes, because the log volume is too large, vector can no longer get up, always oom. I expanded from 2 to 5, but it still doesn't work. Only another consumer group can continue to work. What should I do about it?I can't extend vector indefinitely