node.js --trace-gc 输出的含义是什么？

Question

当使用 --trace-gc 选项时，我得到低于 node.js 的输出

[24851]      852 ms: Scavenge 26.8 (57.2) -> 20.8 (58.2) MB, 3.1 ms [allocation failure].
[24851]      927 ms: Scavenge 28.7 (58.2) -> 22.4 (59.2) MB, 2.0 ms [allocation failure].
[24851]     1013 ms: Scavenge 30.4 (61.2) -> 24.1 (61.2) MB, 2.3 ms [allocation failure].
[24851]     1067 ms: Scavenge 31.6 (62.2) -> 24.8 (62.2) MB, 2.1 ms [allocation failure].
[24851] Speed up marking because marker was not keeping up
[24851] Postponing speeding up marking until marking starts
[24851] Speed up marking because marker was not keeping up
[24851] Postponing speeding up marking until marking starts
[24851] Speed up marking because marker was not keeping up
[24851] Postponing speeding up marking until marking starts
[24851]     1359 ms: Mark-sweep 90.9 (115.1) -> 53.6 (113.7) MB, 9.0 ms (+ 19.3 ms in 105 steps since start of marking, biggest step 1.7 ms) [GC interrupt] [GC in old space requested].
[24851]     1578 ms: Scavenge 69.6 (113.7) -> 62.4 (115.7) MB, 6.0 ms [allocation failure].
[24851]     1676 ms: Scavenge 73.1 (117.7) -> 64.1 (121.7) MB, 5.9 ms [allocation failure].

有人可以解释上面输出中每个项目的含义吗？

Answer 1

我也没有看到任何关于此的文档，但是查看process.memoryusage() 的结果以及使用--trace_gc 选项表明：

Scavenge <heapUsed before> (<heapTotal before>) -> <heapUsed after> (<heapTotal after>) MB, <how long the GC pause was>

heapUsed 和 heapTotal 在哪里被 process.memoryusage() 记录，但更好地解释 by this answer。

Answer 2

当您使用垃圾收集器监控内存分配和CPU利用率时，您需要使用node --trace_gc server.js启动服务器，您需要继续Activity Monitor => Memory => Search node，您可以看到实际使用了多少内存节点处理，现在当您触发上述命令时，您可以在控制台中看到许多日志。从这些日志中主要需要注意两个因素，即Scavenge 和Mark-sweep。如果它只在日志中处理Scavenge，则意味着它是正常的内存使用......但是当Mark-sweep 出现时意味着一些CPU 密集型任务正在那里运行，这会导致高CPU 使用率和内存分配。基本上，这发生在您执行一些阻塞当前堆栈的代码时，该堆栈将在同步过程完成后从堆栈中弹出。为了避免Mark-sweep，您需要使您的函数异步。我希望你能明白我的意思！