On Tue, Aug 26, 2014 at 11:42 PM, Douglas A. Augusto <daaugu...@gmail.com> wrote: > On 25/08/2014 at 21:41, : > I'd like, though, to propose some modifications to the script. Here is the > modified script: > > cat original.slf > updated.slf
'cat' is bad. See below. > nice parallel --timeout 10 --nonall -j0 -k --slf original.slf --tag > echo | sed -e 's/\t$//' -e '/\t/d' > tmp.slf Maybe --timeout 1000% would make more sense: We do not know if these machines are spread all over the world (or if it is done through Tor) and whether 10 seconds is enough, but I would imagine that 10 times the median value would always be enough. > cat tmp.slf > updated.slf 'cat' is not atomic, so you will be inviting a race condition, where the 'cat' runs exactly when GNU Parallel reads the file. 'mv' is atomic - use that instead. Think of how you can automatically start the script when GNU Parallel is started, and stop when GNU Parallel stops. Maybe something like: parallel_daemon start --slf original.slf parallel --slf $UPDATED_SLF ... parallel_daemon stop Or even: parallel_with_daemon --slf original.slf [... other parallel options ...] How is cleanup done? How is cleanup done if the user presses Ctrl-C? What happens if the users run multiple instances? With different original.slf? /Ole
