Hi, This is not blocking my workers at all. The issue it was causing me is I had another process node running on main thread expecting to be called every half a second and sometimes this stall on main thread stopped this from happening.
The thing with allocation is I'm not really allocating memory in data path but I do allocate massive chunks in my init functions. The general description of what I'm doing is keeping the indices of packets buffered somewhere and this is done in a per session manner, for every session at some point it's destination will be decided and all the buffered packets of that session alongside with any future packets of that session will be enqueued to that destination node. I simply keep an array of buffer indices for every session (I'm using a pool so no allocation in hot path) and when decided enqueue it to needed destination. what this causes is at some loops I'm enqueuing more packets than what has been recieved in the same loop(I assume this is OK). Generally I believe the issue is more because of massive amount of memory firstly in buffers that are allocated on hugepages and also the arrays I keep indices in. All things considered I really don't thinks it's a great idea for main to take this long updating statistics just because the program is using a lot of memory. Best regards Parsa
-=-=-=-=-=-=-=-=-=-=-=- Links: You receive all messages sent to this group. View/Reply Online (#26569): https://lists.fd.io/g/vpp-dev/message/26569 Mute This Topic: https://lists.fd.io/mt/116484636/21656 Group Owner: [email protected] Unsubscribe: https://lists.fd.io/g/vpp-dev/leave/14379924/21656/631435203/xyzzy [[email protected]] -=-=-=-=-=-=-=-=-=-=-=-
