Open laggingreflex opened 8 years ago
v8 GC strategy depends on amount of available memory. Simple check of occupied process memory can not say that it's a real memory leak. Because node's memory really increase to some point after run.
May be, you really have few of memory, and forked phantoms leave nothing for node.js?
I'm creating lots of browser instances with
driver.create()
in a batch (~10 or so instances at once) and even though I'm exiting them all withbrowser.exit()
before creating another batch, I'm seeing huge spikes of memory and eventually the node process exits giving memory allocation error (in pm2).I measured by simply creating new instances and calculating increase in
process.memoryUsage().heapUsed
and it shows ~1-3% increase on every instance, eventually building up ~45% in just 20 iterations (create=>exit=>create).Could this be a bug in the library or is this bound to happen when you create so many instances?
edit: added a test