currently, function-mesh uses memory requests for JVM and pass the value as -Xmx, when the JVM reaches the maximum memory usage, it will throw OOM error and crush the POD. A common way to prevent such case is to add a 10% padding to memory limits, to prevent the POD OOM. But for some cloud vendor, it will be needed to make the requests and limits to be equal, so we will need to make the requests and limits as a padding contained value, and use the 90% value instead.
currently, function-mesh uses memory
requests
for JVM and pass the value as-Xmx
, when the JVM reaches the maximum memory usage, it will throw OOM error and crush the POD. A common way to prevent such case is to add a 10% padding to memorylimits
, to prevent the POD OOM. But for some cloud vendor, it will be needed to make therequests
andlimits
to be equal, so we will need to make therequests
andlimits
as apadding
contained value, and use the90%
value instead.