You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
We run a fairly aggressive distributed selenium grid, executing ~200 tests per node machine per hour. Each test is run using a new chrome driver instance for each test, and properly calling driver.quit() at the end of each test. We have found that the memory on our nodes gradually fills up. After some investigation, we found that our selenium standalone server was holding on to thousands of handles to dead chromedriver instances. Each one was taking up 20kb of ram, and no longer attached to any live process / PID. After resetting the server, these zombie processes were immediately released from memory. At present we have not attempted to replicate the issue using firefox/IE or through local execution.