Fetcher cache is great for not re-downloading the binary, but you might also consider if you just want a long-lived executor that survives without any tasks, awaiting new tasks to eventually come its way. That approach uses up some resources while the executor idles, but could dramatically decrease task launch time, if you're just spawning a thread instead of a container/process.
On Wed, Jun 29, 2016 at 6:45 AM, Tomek Janiszewski <[email protected]> wrote: > You can use Fetching through the cache > <http://mesos.apache.org/documentation/latest/fetcher/> to download file > only once. > > śr., 29.06.2016 o 15:19 użytkownik Scott Weiss <[email protected]> > napisał: > >> Every time my framework launches a task, it downloads the (custom) >> executor from the URI I provide. then when the task finishes executing, the >> framework gets the message Executor lost. this seems to cause the slaves to >> download the executor binary every time they launch a task. Is there a way >> to keep the executor binary on the slaves so it only has to be downloaded >> one time? >> >
