Deadline Slave memory leak?

Deadline
1

Hi we have more and more slaves displaying this behavior. The slave stall, when we go on the machine to see what is going on, we see DeadlineSlave.exe using up insane amount of RAM. I don’t think the slave should use up that much ram! So i’m suspecting a memory leak.

FYI. These machines are running window XP pro 64 bit. They have 12Gig of physical RAM

deadline_memory_usage.jpg
deadline_memory_usage.jpg751×720 183 KB

2

That’s strange. Almost our entire farm is XP 64, and we haven’t seen this behavior. Any additional information you can pass along would be great, such as:

  1. On average, how long is a slave running before it shows this behavior?
  2. Does the memory climb gradually after the slave is first launched, or is there a sudden memory “explosion”, at which point the slave crashes?
  3. Does this happen if a slave just sits idle and doesn’t render any jobs? To test this, you can unassign the pools and groups from the slave and ensure that no jobs are submitted to both the ‘none’ pool and ‘none’ group.

Anything else that you think would be helpful would be great!

Cheers,

  • Ryan
3

We did more digging and we found out that this memory leak appends after the slave crashes. Window display the we are sorry your software crashed, then the slave stay open until I press the ok button of the windows warning… during all this time, the slave memory keeps on climbing.

So the leak seems to be windows related…not deadline related.

We just need to figure out why the 3.0 slave are crashing so often. I might have to try the latest beta on the entire farm to see if I get fewer crashes