Long Running Processes: Unterschied zwischen den Versionen

Aktuelle Version vom 19. März 2018, 07:37 Uhr

All our interactive login machines

...are not really meant to be used for long running computations.

The might reboot over night without further notice.

@@ Zeile 1: / Zeile 1: @@
+All our interactive login machines
-== The problem ==
+* the Pool computers
-From time to time we see applications running for a long time - days, not hours. Most times this is a crashed process or a bad programmed application which refrains to die  ;-)
+* the old login-VMs
+* the newer [[Shell]]-VMs
-On the other hand there may be tasks that simply need so much time that they run through the weekend to solve a problem. (Matlab problem solvers or simulations for example.)
+...are not really meant to be used for long running computations.
-We (the system administrators) usually have no clue which case is true for a specific program. Should we kill the processes to clean up the system? Or should we leave it running?
-To emphasize it: we have nothing against running programs a long while. If you want to be friendly you should run those program using "nice" though.
-== The solution ==
-To avoid getting killed '''please enter your expected usage''' by adding a line to this table:
- Username      Machine  Application    Expected end of execution
- johndoe       c099     myapp.java     31.12.2099
-''(This is a wiki - click on "Edit" after "Login" in the top right corner)''
-Running processes without "nice" and without being listed above will get killed without further notice.
-This problem is relevant for the physical pool clients and for the virtual login [[Shell]]-machines.
+The might reboot over night without further notice.